YC Winter 2025 (W25) companies

a0.dev builds an AI-powered platform that lets mobile developers create, publish, and monetize React Native apps in days rather than weeks. Developers describe what they want in plain language, and the platform's AI agent writes and edits the code in real time, then handles one-click deployment to the App Store and Google Play. It replaces the manual coding, backend setup, and store-submission work that traditionally stretches a mobile build cycle across weeks.

A1Base builds communication infrastructure for AI agents, giving them the ability to send and receive messages across email, SMS, WhatsApp, Slack, and other channels. It targets developers and AI product teams who need their agents to interact with end users through existing messaging platforms. The product replaces custom-built integrations for each channel.

Adam builds an AI-powered CAD workspace for hardware engineering teams. It converts text prompts into parametric designs and handles surrounding engineering work: supplier research, BOM cleanup, component sourcing, factory handoff packaging, and design review preparation. It replaces the manual coordination and documentation load that slows hardware teams between a design decision and a shipped part.

AfterQuery builds expert-curated training datasets for frontier AI model development. It captures how domain professionals reason through decisions, tradeoffs, and context, then structures that thinking into supervised fine-tuning data, reinforcement learning rubrics, agent environments, and computer-use trajectories. The product replaces lower-quality, output-only data pipelines that cause model performance to plateau on complex, real-world tasks.

Agentin AI builds AI agents that automate processes across enterprise software systems such as Salesforce, NetSuite, and SAP. The agents move data and execute actions across these platforms, replacing manual work that today requires human operators to navigate heavily customized enterprise environments. Agentin AI trains its agents using proprietary reinforcement learning techniques so they adapt from failures and handle the unique configurations each enterprise deploys.

Alice.tech converts students' own course materials, lectures, and textbooks into structured study plans, flashcards, practice questions, and mock exams. It targets college and high school students who would otherwise review slides manually or rely on generic study tools. The product replaces unguided self-study with an adaptive, course-specific preparation loop built around the student's actual syllabus.

Altrina builds AI agents that execute standard operating procedures across browser and computer interfaces, replacing manual, rules-bound workflows that staff would otherwise perform step by step. The platform has been deployed in local governments, law offices, insurance brokerages, consulting firms, and hedge funds. Altrina has since been acquired by Abridge, where the team continues work on AI automation in healthcare.

Amby Health builds AI software for ambulance agencies that automates billing and clinical quality review. It analyzes patient care reports to code insurance claims and flag liability and care risks, replacing the back-office teams that agencies currently use to do that work manually. The process that once took months runs in seconds.

Archon helps software companies break into federal government sales by accelerating the FedRAMP compliance process, cutting the authorization timeline from roughly 16 months down to 6 and reducing the associated cost by around $1 million per company. FedRAMP certification is a mandatory security requirement for cloud software sold to U.S. federal agencies, and the process traditionally demands significant time and specialist resources. Archon replaces that slow, expensive compliance grind with a faster, cheaper path to government contracts.

Artifact builds a collaborative, version-controlled ECAD tool for electrical system design on complex hardware programs. Hardware engineering teams use it to design wiring harnesses, trace signals across subsystems, and auto-generate harness drawings, pin-tables, and bills of materials from a single source of truth. It replaces fragmented spreadsheets and tribal knowledge handoffs across design, procurement, and manufacturing teams working on aerospace, defense, and other technically demanding systems.

Artificial Societies builds AI-powered simulations of large human populations, letting companies test marketing, content, and product decisions against a synthetic model of their target audience before real-world launch. It replaces traditional market research, which captures isolated individual responses and misses how people influence one another at scale. Clients use it to run scenario testing, crisis and communications strategy, and market expansion analysis across simulated networks of consumers, investors, and regulators.

assistant-ui builds an open-source TypeScript/React library that gives frontend developers production-ready chat UI components and a runtime for AI agents. It handles streaming, tool calls, branching, and generative UI, replacing the work of building these primitives from scratch. Integrations with LangGraph, LangChain, and the Vercel AI SDK let developers connect any backend through adapters.

Asteroid builds and hosts AI browser agents that handle back-office operations for enterprise customers, including those in regulated industries such as healthcare. The agents navigate web portals to execute tasks like EHR data extraction, prior authorization submission, eligibility verification, and provider enrollment, replacing manual staff work across these workflows. Companies can either build agents on Asteroid's platform themselves or have Asteroid build and maintain them.

Astro develops renewable energy projects from the ground up, using AI models to identify viable grid connections, acquire land, and secure interconnection agreements before selling the completed development packages to large energy companies. The work it replaces is the manual, high-failure-rate site selection and grid study process that causes the vast majority of renewable projects to collapse on unexpected interconnection costs. Customers are large energy developers and utilities that buy shovel-ready projects rather than originate them internally.

AthenaHQ builds a generative engine optimization platform that helps B2B marketing teams get their brands surfaced by AI search tools such as ChatGPT, Perplexity, Claude, and Gemini. The platform tracks brand visibility across more than eleven large language models, surfaces content optimization recommendations, and analyzes citation sources. It replaces manual, fragmented tracking workflows with a single command center for AI search strategy.

awen builds a voice-driven visual creation platform for creative teams, replacing the manual prompt engineering and parameter tuning required by traditional AI image and video tools. Users describe what they want in plain language, and awen selects the right models, handles orchestration, and returns images, videos, and visual assets, covering use cases from ad campaigns and e-commerce product shots to design exploration and video production.

Axal designs, assembles, and tests custom printed circuit boards for hardware teams. The company handles the full PCB pipeline, from schematic design and component sourcing through in-house assembly, X-ray and electrical testing, and delivery with build documentation. This replaces the fragmented process of coordinating separate design firms, contract manufacturers, and external assemblers.

Axiom builds a decentralized finance trading platform for retail crypto traders. It covers memecoins, perpetuals, and yield products in a single interface, replacing the need to navigate multiple DeFi protocols separately.

Bezel builds an AI platform that generates product photos and videos of clothing on virtual human models for fashion e-commerce brands. Brands upload garment images, choose a model, and receive studio-quality visual assets without organizing a physical shoot. The product replaces traditional photography production, including casting, scheduling, and on-set execution.

Bild AI builds takeoff and estimating software for Division 8 distributors and estimators, the specialists who quote doors, frames, and hardware on commercial construction projects. It reads uploaded blueprint packages, extracts door schedules and hardware specs, sizes components by opening dimensions, and maps approved manufacturer alternates, replacing the manual plan-reading and spreadsheet work estimators currently do to produce a bid.

Bindwell builds AI-driven systems to discover and develop new pesticides in-house. The company pairs scalable wet labs that generate molecular data rapidly with neural networks trained on that data, replacing the slow, manual cycle of traditional pesticide and molecular design. A candidate molecule goes from design to a real biological result in a day, a feedback loop the company uses to iterate toward safer, more effective compounds.

BlindPay builds a stablecoin API that lets companies send and receive money across borders using fiat currencies, stablecoins, and multiple blockchains. It handles compliance and regulatory requirements on behalf of its customers, replacing the need for businesses to build or manage that infrastructure themselves.

Bluebook builds AI agents for accounting firms, handling research, reconciliations, and monthly closings. The product targets the bottleneck of manual, time-intensive work that caps how many clients a single accountant can manage, replacing rule-based software and hours of manual review. Firms gain the capacity to take on a much larger client load while shifting accountant time toward advisory work.

Browser Use builds open-source web agents and remote browser infrastructure that let AI models control a browser to complete tasks autonomously. Developers and teams use it to replace manual, browser-based workflows, from navigating protected sites to executing multi-step web tasks. The core library, one of GitHub's most-starred projects, ships under the MIT license and can run in the cloud or self-hosted.

Butter builds an LLM proxy that records how an agent completes a task and then replays that exact sequence deterministically, bypassing the model entirely on repeat runs. The target customer is any team running AI agents on repetitive workflows where the same tool calls fire again and again. The product replaces live model inference for those cases, cutting latency, token spend, and output variability.

Bystreet builds an AI-powered database of hotel ownership data for hospitality firms that need to reach property owners and decision-makers. It covers hundreds of thousands of hotels globally, providing verified beneficial owner identities, key contacts, and direct contact details. The product replaces manual research and fragmented data sources that teams use to identify and approach hotel owners.

Calltree builds AI support representatives for enterprise call centers, handling both voice calls and chat. Where most AI agents are limited to scripted flows, Calltree's agents connect to internal tools, including proprietary CRMs and operational dashboards, doing the same work a trained human agent would. The target customer is enterprises running large support operations who want to automate frontline interactions without sacrificing access to their existing systems.

Candid builds a hiring platform aimed at connecting businesses with high-quality talent. It targets B2B companies looking to fill roles, replacing manual or fragmented recruiting workflows with a structured screening and matching process.

Candor builds an AI-powered platform that helps defense and deeptech startups secure government funding. AI agents surface relevant procurement opportunities and handle the application work that previously took weeks, compressing it to days. The product gives small companies the same access to government contracts that large, established defense contractors have long held.

Cardamon builds an AI platform that automates regulatory compliance for financial firms. Its agents ingest regulations across jurisdictions and languages, extract the obligations that apply to a specific firm, map those obligations against existing controls, and surface gaps with suggested fixes. This replaces weeks of manual work by compliance officers translating dense regulatory text into product requirements.

careCycle builds an AI-powered operating system for Medicare and ACA agencies, replacing manual agent workflows across the full member lifecycle from lead intake through enrollment and ongoing retention. Coordinated teams of voice AI agents handle after-hours calls, scheduling, qualification, and warm transfers to licensed agents, cutting the post-enrollment disengagement that drives chronic member churn. The platform also serves as a unified CRM, giving agencies a single system of record in place of disconnected tools.

Caseflood.ai builds an AI-powered intake platform for law firms, handling the full inbound pipeline from first contact through qualification and signing. It replaces the manual work of intake coordinators and operations staff: reaching leads faster, guiding reps in real time, and ensuring no prospect falls through the cracks. The target customer is the ambitious plaintiff-side firm that wants to sign more cases without scaling headcount proportionally.

Cedar builds an open-source React framework that lets developers embed AI agents directly into their application's frontend, moving user interaction beyond a chat window. Product teams use it to add voice interfaces, human-in-the-loop controls, and real-time Slack handoff to their existing apps. It replaces custom glue code and reduces reliance on customer success managers to walk users through complex workflows.

Cenote builds HIPAA-compliant AI sales agents for telehealth and direct-to-consumer health clinics. The agents handle patient conversion and reactivation, replacing generic broadcast texts and manual human follow-up for abandoned checkouts and lapsed customers. They engage leads immediately across call, SMS, and WhatsApp, with responses tailored to each patient's history and schedule.

Cifrato builds AI agents that run end-to-end accounting workflows for businesses, replacing manual data entry and bookkeeping staff. The agents pull invoices directly from tax authorities and email, classify and reconcile each document against the correct accounts, post finished entries into existing ERPs, and handle tax filings. Finance teams receive only the exceptions that require human judgment, plus on-demand CFO-level reporting from the closed books.

Closure builds an AI-powered analysis platform for law enforcement investigators, district attorneys, and their staff. It ingests large volumes of unstructured evidence, including recorded calls and digital data, and lets investigators search, query, and surface relevant details in minutes rather than hours or days. The work it replaces is manual review: combing through recordings, transcripts, and case files by hand to find witnesses, inconsistencies, and critical information.

Confident AI builds an evaluation and observability platform for teams shipping LLM-powered applications. Engineers, product managers, QA teams, and domain experts use it to benchmark model quality, monitor production behavior, red-team against adversarial inputs, and enforce AI governance standards across an organization. It replaces ad hoc, team-by-team quality checks with a shared evaluation layer built on DeepEval, the company's open-source LLM testing framework.

Conntour builds an AI platform that lets governments and corporations query large networks of security cameras in plain language. Operators can search historical footage, set custom event alerts, and pull aggregate data, replacing manual footage review and the rigid, pre-defined filters of legacy video surveillance systems.

Contrario builds an AI recruiting platform that pairs artificial intelligence with human expert recruiters to fill critical roles for B2B companies at every stage, from pre-seed through post-IPO. It handles sourcing, screening, and pipeline management across engineering, GTM, operations, and product design, replacing the work of in-house recruiting coordinators and traditional staffing agencies.

CopyCat builds an AI platform for dermatology clinics that handles biopsy result callbacks: a voice AI contacts patients, explains their results in plain language, answers common questions, and books follow-up treatment appointments. This replaces the manual phone outreach that clinic staff would otherwise make for each lab result. The program description establishes the dermatology use case; the website describes an unrelated insurance proposal tool for brokers, and the two cannot be reconciled, so the profile relies on the program description.

Cuckoo Labs builds a real-time AI translation tool for B2B sales, marketing, and support teams conducting multilingual meetings. It joins Zoom, Google Meet, Slack, and Microsoft Teams calls, ingests company-specific keywords and documents, and delivers live translation across 20-plus languages tuned to technical and industry vocabulary. It replaces human interpreters and post-meeting translation work for global customer-facing conversations.

Dalus builds model-based systems engineering (MBSE) software for hardware teams in aerospace, defense, robotics, energy, and automotive. Engineers use it to model system architecture, manage and trace requirements, run analyses, and verify designs in a shared environment built on the SysML v2 specification. It replaces fragmented documentation, siloed toolchains, and manual requirements tracking that cause misalignment and costly late-stage design changes.

Dartboard Energy builds an AI analyst for electricity markets, serving owners of grid-scale battery storage projects. It processes market data to support the decisions those owners currently handle through manual analysis or generalist tools. The product targets a technical, commercially sophisticated buyer who needs continuous visibility into power market dynamics to operate and optimize large storage assets.

Delineate builds AI agents that design clinical trials for pharmaceutical companies. The agents process existing trial research into specialized datasets, replacing months of manual work by scientists. The result is faster trial design and progression through trial phases.

Demeter builds an AI platform for private equity firms to manage and grow portfolio value. It pulls in deep research to enrich portfolio company data, then automates performance tracking and benchmarking, replacing the manual analyst work that value creation teams do across spreadsheets and fragmented sources.

Dollyglot builds a conversational AI platform where users speak with photorealistic, real-time video avatars of real or fictional personas. A user submits a photo and voice sample to generate an interactive character, replacing static chatbot text exchanges with live video conversation. The product targets consumers who want immersive, character-driven interaction rather than conventional text-based AI chat.

Edexia builds an AI grading assistant for essay-based subjects, aimed at secondary school teachers working across curricula such as IB, VCE, HSC, QCE, and WACE. Teachers train the tool against their own rubrics and review all feedback before it reaches students, replacing the manual work of drafting written comments and scores. The product learns from teacher corrections so its judgments align with each educator's standards over time.

Egress Health builds AI agents that automate revenue cycle management for dental offices. The platform handles eligibility checks, claims scrubbing, payment posting, and denial management, replacing the manual billing work typically done by in-house administrative staff.

Enhanced Radar builds applied AI systems for aviation, targeting safety and efficiency in air traffic control. Its core product is an aviation-specific AI trained on pilot-controller communications datasets that Enhanced Radar collects, labels, and quality-checks in-house. The work it replaces is manual, error-prone monitoring of complex airspace operations, including real-time detection of runway incursion risks.

Ergo builds AI agents for B2B sales teams that monitor customer communications across Zoom, email, and Slack, then automatically update CRMs like HubSpot and Salesforce. The agents also trigger and execute mid-funnel follow-ups based on market signals and past interactions, replacing the manual logging, pipeline hygiene, and outreach sequencing that sales reps and revenue operations teams otherwise do by hand.

Excellence Learning builds an AI-powered math and science tutor for students and teachers, targeting US parents as the primary buyers while also selling to schools. For students, it replaces private tutors and structured drill programs like Kumon by adapting exercises, explanations, and study plans to each student's knowledge level and learning style. For teachers, it automates assignment creation, exam generation, and worksheet grading, replacing hours of manual preparation and correction.

Exin Therapeutics builds an AI-driven drug discovery platform for neurotherapeutics, targeting diseases defined by dysfunctional neural activity. The platform uses multimodal AI models trained on high-density in-house mouse data to predict how drug candidates affect multiple neurological disorders simultaneously, starting with epilepsies linked to autism and Parkinson's disease. It replaces conventional trial-and-error screening by directing target discovery across disorders and therapeutic modalities from a single platform.

Exla builds an SDK that compresses and accelerates transformer models, including LLMs, VLMs, and VLAs, for deployment across any hardware. It cuts model memory footprint by up to 80% and speeds up inference by 3 to 20 times through aggressive quantization, replacing custom optimization work that engineering teams would otherwise do by hand. The integration requires only a few lines of code.

finbar builds an AI investment research platform for investors, analysts, and quantitative developers. It delivers company primers, prebuilt financial models, AI-assisted document analysis, and fundamental data, replacing the manual work of building models from scratch and combing through source filings. Hedge funds and analysts use it to move from raw data to an investment thesis without the time-consuming groundwork.

Fira builds an AI research platform for investment analysts, replacing manual work across SEC filings, earnings reports, and internal documents. Analysts ask questions across hundreds of financial documents and get source-cited answers, verified calculations, and outputs ready to export to Excel. It handles trading comps, EV bridges, number verification, and multi-document synthesis in a single workspace.

Flowtel builds an AI operating layer for hotels, handling guest calls and property operations through voice agents. The platform covers tasks from booking to room service, replacing the mix of front-desk staff and fragmented tools that hotels rely on today. A single live surface manages all properties, with custom AI deployed in production within 30 days.

Forge Automation manufactures low-volume custom CNC metal parts for mechanical engineers, shipping orders in four days or less. Engineers upload a CAD model to the platform and receive finished parts roughly three to four times faster than the two-to-three-week lead times common at traditional suppliers. The service replaces the slow quoting, scheduling, and production cycles that typically stall hardware development teams.

FuseAI builds an agentic outbound sales platform for B2B sales and marketing teams at startups and enterprises. Its AI agents handle contact enrichment across 40-plus data providers, lead identification based on buying signals, and multichannel outreach via email, LinkedIn, and phone, replacing the patchwork of point solutions a sales rep would otherwise juggle. The platform consolidates prospecting, qualification, and engagement into a single interface, including a Claude integration that runs sales workflows directly inside that assistant.

G LNK builds a healthcare commercial intelligence platform that connects life sciences companies with medical professionals for drug development and commercialization. It uses proprietary matching and ranking models to surface qualified clinicians, drawing on a database of HCP profiles, prescribing behavior, claims data, payments, and fair market value benchmarks. The platform replaces manual expert identification and outreach while keeping engagements within regulatory and ethical compliance requirements.

Gale automates U.S. business immigration for companies, handling visa applications and compliance workflows that HR and legal teams otherwise manage manually. It coordinates with independent attorneys to deliver legal services, replacing the slow, fragmented process of juggling outside counsel and internal paperwork. The platform also connects with existing HR systems to keep employee immigration status in sync.

General Trajectory builds AI systems for physical-world applications, targeting autonomous defense operations and scientific research and development. The work it replaces is the manual planning, analysis, and decision-making that currently limits speed and scale in those domains.

GradeWiz builds an AI grading assistant for K-12 and higher-education teachers. It generates rubrics, scores handwritten worksheets and tests, and returns detailed written feedback to students, replacing the manual grading work educators do outside the classroom.

Haleum builds an AI-native data loss prevention and insider risk platform for enterprise security teams. Its product, Candor, monitors communication channels and user activity across an organization, then uses AI agents to assemble the full context behind each event and surface only the cases that genuinely warrant action. It replaces static detection rules and manual alert triage, catching insider threats, IP theft, compliance violations, and shadow AI data exposure that rule-based DLP tools miss.

Harbera builds an AI-native revenue cycle platform for clinic and health system administrative teams, covering provider credentialing and fee schedule automation. It replaces manual work across document ingestion, payer enrollment application prep, and follow-up tracking by continuously monitoring provider credential status and flagging expirations before they cause denied claims or out-of-network gaps. The platform consolidates fee schedules, credentialing workflows, payer communications, and reporting into a single operating view.

Harper builds an AI-native commercial insurance brokerage for mid-sized businesses with complex coverage needs. Brokers backed by an AI system shop every relevant market, including specialty carriers most generalist agents cannot access, and return fully explained options to the client. The product replaces the fragmented work of sourcing, comparing, and binding commercial policies across multiple insurers.

HealthKey builds an AI-powered patient identification platform for clinical trial sites. It connects to a clinic's existing electronic health records, encodes each trial's eligibility criteria, and automatically prescreens patients against those criteria, replacing the manual chart reviews that coordinators do today. The result is a ranked, evidence-backed list of candidates delivered to research teams, so doctors' offices can enroll more patients and capture the per-enrollment fees that pharmaceutical sponsors pay.

Hey Telo builds voice AI for home services businesses, handling inbound and outbound calls that would otherwise require a human dispatcher or receptionist. It targets the home services sector, where missed or mishandled calls directly cost jobs and revenue.

hud builds evaluation and reinforcement learning environment infrastructure for computer use agents, the AI systems that autonomously browse the web. Its customers are frontier AI labs that need to measure and improve agent reliability across a wide range of real-world tasks. hud replaces ad hoc, internally built eval tooling with a structured framework purpose-built for web-browsing agents.

Infinite builds a stablecoin-based payment infrastructure for businesses that need to move money across borders. Its platform combines global USD and multicurrency accounts with AI-native compliance tooling, replacing the patchwork of traditional correspondent banking, manual KYB workflows, and expensive cross-border wire infrastructure. Platforms and businesses embed Infinite's APIs and SDKs to launch branded accounts, run payouts and payins, and handle regulatory compliance without building those systems themselves.

Instinct builds a small lunar lander designed to reach the Moon's surface via standard rideshare launches to low Earth orbit, eliminating the need for dedicated heavy-lift missions, in-orbit staging, or refueling. The lander delivers at least 20 kg of payload to the lunar surface and operates throughout a full lunar day. It cuts the cost and complexity that currently make Moon missions rare and slow.

Inversion Semiconductor builds lithography machines for advanced chip fabrication. Its core innovation is a miniaturized particle accelerator, shrunk by a factor of one thousand, that generates a high-power light source to pattern transistors at smaller scales than current equipment allows. The machine doubles transistor density and triples throughput, replacing existing lithography tools for chipmakers pushing transistors to their physical limits.

Invo builds an AI automation suite for food distributors and manufacturers, handling purchase order processing, RFQ evaluation, and accounts receivable collections. It replaces manual back-office work: clerks entering POs from emails and PDFs, staff reviewing and quoting RFQs by hand, and AR teams sending payment reminders one by one. The platform connects to existing ERPs and runs automated outreach, anomaly detection, and predictive payment analytics in their place.

Karoo builds a fully online consumer lending product for Canadians, offering personal loans between $500 and $5,000 with instant pre-approval decisions and same-day funding. It targets borrowers shut out of or underserved by Canada's incumbent bank branches, replacing a slow, in-person loan process with a transparent, digital one that carries no hidden fees and no early repayment penalties.

Karsa builds business and personal banking accounts for people in emerging markets, where reliable access to dollar-denominated accounts is scarce. Customers hold multi-currency balances, receive payments from platforms like Shopify, Stripe, and Upwork, and send money to suppliers and contractors across 80-plus countries via ACH, SWIFT, SEPA, and stablecoin transfers. The product replaces fragmented workarounds that cross-border freelancers, e-commerce resellers, and small businesses currently use to get paid and pay out in hard currency.

Leaping AI builds voice and text agents that handle inbound and outbound calls for mid-market and enterprise contact centers, replacing the work done by human agents in customer support, lead qualification, and appointment scheduling. Target verticals include home services, real estate, health insurance, Medicare, travel, and retail. The platform connects to CRMs and other business systems to conduct full conversations autonomously, end to end.

Lightbox gives small businesses an AI-powered graphic design tool to build a brand identity without hiring a professional designer. It targets owners who need logos, visual assets, and cohesive brand materials but cannot afford agency or freelance rates. The product replaces the traditional workflow of briefing, iterating with, and paying a human designer.

Lopus AI builds a unified analytics platform for go-to-market teams at B2B companies. Its product, Lopus Probe, connects CRM, billing, product usage, call transcripts, and messaging data into a single governed workspace, then uses an agentic semantic layer to answer business questions with full source lineage. It replaces the manual work of data teams fielding ad-hoc requests and revenue teams stitching together context across disconnected tools.

Lucid builds neural world models that simulate interactive, photorealistic environments, replacing traditional hard-coded game and physics engines with video-generating neural networks. The models are action-conditioned: a user's inputs drive what the simulation renders next, in real time. Lucid's current architecture uses aggressive tokenizer compression to run at high frame rates on consumer-grade GPUs, with far lower training compute than comparable systems.

Lucidic AI builds a continuous testing and optimization platform for enterprise AI agents. It ingests a company's production logs, edge cases, and operational policies, then runs controlled simulations and automated optimization to find failure modes and apply fixes before they reach production. The result replaces manual prompt engineering and ad-hoc debugging with a closed-loop system that keeps agents aligned to a company's specific domain logic and rules.

Macadamia builds AI agents for engineering firms in mechanical engineering, manufacturing, and construction. The agents handle routine technical and administrative work, including CAD drafting, P&ID generation, QA inspection, cost estimating, and data entry. They connect directly to existing engineering data sources, technical documentation, and third-party tools, replacing the low-leverage tasks that otherwise fall to human technicians and back-office staff.

Maive builds an AI sales platform for home services contractors, sold under the name Ladder. It coaches sales reps by analyzing their calls, automates follow-ups and review requests, and handles the administrative work that currently falls on reps after each appointment. The product targets home services enterprises whose field reps sell projects door-to-door or on-site.

Maritime Fusion builds compact tokamak fusion reactors designed for large commercial vessels and off-grid installations, replacing diesel and alternative marine fuels with zero-emission power. By targeting applications that need roughly one-tenth the output of a grid-scale plant, the company sidesteps unsolved material-science challenges that make early fusion economically unviable on the grid. While the reactor advances, Maritime Fusion is commercializing high-temperature superconducting cable technology for power distribution as an intermediate product.

Mastra builds an open-source TypeScript framework that lets product developers construct, deploy, and evaluate production AI agents and applications. It ships primitives for multi-step workflows, autonomous agents, retrieval-augmented generation, and output evaluation, replacing the need to wire these components together from scratch. The target user is a software developer who needs to ship agent-based features, including human-in-the-loop workflows and long-running agents, without building core infrastructure by hand.

Mecha Health builds foundation models that automate the analysis of X-rays for radiologists, turning medical images into complete draft radiology reports. The models replace the manual read-and-dictate workflow that currently limits a radiologist to roughly one scan per hour. Mecha Health's models are trained on significantly less data and at a fraction of the size of comparable models from major AI labs, while outperforming them on clinical accuracy benchmarks.

Mentra builds an open-source operating system and app store for smart glasses, targeting frontline and physical workers. Its flagship product, Mentra Live, gives field technicians hands-free access to point-of-view video calling, AI guidance, and automatic work documentation, replacing manual logging, separate devices, and delayed expert support. The platform runs on customer-controlled infrastructure and supports air-gapped deployments, keeping operational data inside the organization.

Mercura builds AI-powered quote and order automation software for distributors and manufacturers in the construction supply chain. It reads incoming customer requests, matches products, and generates quotes, replacing the manual back-office work that sales teams previously handled by hand. Sales reps move from processing requests over days to minutes, and can take on more volume without adding headcount.

Mesh builds accrual automation software for Controllers, Accounting Managers, and CFOs at mid-market companies. It replaces manual spreadsheet-based accrual and prepaid tracking by pulling data from existing procurement, accounts payable, and contract systems to generate consistent, policy-driven journal entries with embedded audit trails. Finance teams close faster and handle only the exceptions that require human judgment.

Misprint builds a combined pricing analytics and trading marketplace for physical collectibles, starting with Pokémon and trading cards. Collectors, investors, and hobbyists currently piece together valuations from auction records, forums, and reseller listings with no single authoritative source. Misprint replaces that process with machine-learning-driven price data and a buy/sell marketplace in one platform.

Miyagi Labs builds an AI-powered exam prep platform for students preparing for high-stakes tests such as the SAT, ACT, NEET, and national exams across Kenya, Nigeria, Turkey, and India. The platform replaces traditional prep books and human tutors with an AI tutor that delivers Socratic explanations, adaptive practice drills, diagnostic score predictions, and full-length timed mock exams. Students get a continuously updated study path that targets their weakest skills rather than working through static, finite question sets.

Mosaic builds an agentic video editing platform that lets users construct and run custom AI agents on a node-based canvas. Editors and creators use it to automate multi-step video editing workflows that would otherwise take hours of manual work. The agents handle the editing process end to end, compressing that effort to seconds.

Mundo AI builds multilingual training data and evaluations for frontier AI labs and AI companies whose models underperform in non-English languages. It supplies audio, video, and emerging-modality datasets along with applied research, replacing the manual, ad-hoc work labs would otherwise do to source and validate non-English data themselves.

Nitrode builds the evaluation, training data, and post-training infrastructure that AI models need to get better at game development. Rather than relying on a single general-purpose model, it develops specialized systems, each measured and improved on a distinct slice of the game-creation process. The work it replaces is the manual, ad-hoc effort studios and AI labs currently spend benchmarking, diagnosing, and improving model performance on game-engine tasks.

Oki builds an AI briefing layer for company leaders. It reads across a team's code commits, project tickets, and internal messages to generate daily, personalized status updates for executives and managers. It replaces the manual work of chasing progress across tools and assembling reports by hand.

Olive builds AI-powered business intelligence tools that let non-technical teams create dashboards and internal data tools by describing what they need in plain language. It targets companies that want production-ready analytics connected directly to their databases or data warehouses, replacing custom BI development work done by engineers or analysts. The product handles schema understanding, access control, and role management so teams can ship and govern data tools without writing code.

Onlook builds a visual, open-source code editor that lets designers and developers build, deploy, and iterate on websites and web apps in one place. It merges design tooling, front-end development, and AI into a single interface, replacing the back-and-forth handoff between design software and code editors. The target user is anyone building for the web who currently splits work across separate design and engineering tools.

OpenIntake builds an AI-powered intake system for law firms that answers inbound calls and qualifies prospective clients around the clock. It replaces the front-desk receptionist and after-hours voicemail that leave callers unattended, converting those calls into signed clients instead of lost leads.

Operand builds an AI platform that replaces management consultants for e-commerce and retail businesses. It ingests data across competitor pricing, ad strategy, and inventory to generate pricing and discount recommendations, work that traditionally required external consulting engagements. Every output is reviewed by former MBB consultants before it reaches the client, combining automated analysis with expert verification.

Optifye.ai builds a computer-vision monitoring system for factory floors, placing cameras at workstations to track operator activity, measure output against standard targets, and flag SOP violations in real time. Operations leaders use the data to identify bottlenecks, compare line performance, and drive continuous improvement. It replaces manual observation and the guesswork that comes from having no structured visibility into individual worker or line productivity.

Orbital Operations builds high-thrust, reusable space vehicles designed to protect critical satellites, including GPS and military communications assets, from adversarial threats. Its flagship platform, Astraeus, uses cryogenic propulsion systems drawn from proven launch vehicle technology to deliver over 100 times the thrust of current in-space propulsion. The vehicles replace the absence of any rapid-response, high-maneuverability option for satellite defense in orbit.

Orchids builds an AI-powered full-stack engineering tool that lets people design, build, and ship apps and websites without traditional development work. It targets anyone who needs to get a product live, replacing the manual coding and configuration that a human engineering team would otherwise handle.

Osmosis builds a forward-deployed reinforcement learning platform that fine-tunes open-source models for specific enterprise tasks, replacing the need for companies to use general-purpose foundation models. Engineering teams use it to train task-specific models for document extraction, AI agent tool use, and code generation, with Osmosis handling the full post-training workflow from reward function design to model serving. The platform monitors production performance and triggers automated retraining runs, keeping models current without manual intervention.

Outlit monitors B2B customer accounts continuously, pulling signals from CRM, product analytics, communication tools, and billing data to surface churn risks and expansion opportunities across the long tail of accounts a sales or customer success team cannot manually review. It replaces periodic manual account reviews with automated detection and action plans executed within controls the team defines.

Overstand Labs builds an on-demand data and AI team for medium-to-large enterprises, unifying communications, such as email, Slack, and calls, with internal operational data into a single queryable environment. It surfaces customer needs, risks, and revenue opportunities that would otherwise require manual analysis or scheduled reports. The work it replaces is a combination of traditional data consulting, bespoke reporting pipelines, and the lag time between raw business data and actionable decisions.

Ovlo runs an AI-native recruiting service for London startups, pairing specialist recruiters with AI sourcing tools to fill technical and niche roles. Founders brief a role, a domain-specific recruiter maps the market and screens candidates, and the startup interviews only pre-vetted people. It replaces slow, generalist recruitment with a faster, transparent process where no search firm work falls on the founding team.

Paragon builds AI agents for industrial distributors and manufacturers, covering order processing, customer engagement, and demand forecasting. Its three products, Forge, Surge, and Alloy, handle the operational work that sales and operations staff currently do manually, from managing seasonal order backlogs to planning inventory. Target verticals include HVAC, pipe and valve, lumber, electrical components, and industrial packaging.

Paratus Health builds AI voice agents that run front-desk operations for outpatient clinics. The agents handle inbound and outbound patient calls, intake, insurance verification, documentation, and billing preparation, replacing the administrative staff who do that work today. Everything connects directly into EHRs such as Epic and Athena, removing the need for separate tools or manual hand-offs between systems.

Pave Robotics builds autonomous robots that seal cracks in asphalt pavement, replacing manual crack-sealing crews for paving companies. The robots operate around the clock and produce more consistent repairs than hand labor, cutting both time and cost on road maintenance jobs.

PAX Markets builds a cryptocurrency centralized exchange where the matching engine and trading logic run on the same physical chip, eliminating the network distance between order origination and execution. High-frequency traders get deterministic nanosecond-level execution through a paid on-chip API, while retail and other traders access the exchange at zero cost with rebates on taking liquidity. The product replaces conventional co-location arrangements, where traders must route orders from external servers to a separate exchange engine over a network link.

Peppr AI builds a real-time AI copilot for B2B sales reps. During live calls, it detects incoming questions, pulls answers from the company knowledge base, and surfaces suggested responses on the rep's screen before they have to stall. It also scores playbook adherence, auto-writes call notes, and converts each conversation into updated battle cards, replacing manual note-taking, ad-hoc coaching, and static sales content.

Permitify builds an AI co-pilot for municipal building plan reviewers, helping understaffed city departments check construction documents for code compliance. The tool lets a single reviewer cover the work that would otherwise require far more staff, cutting the time between permit application and approval. It replaces manual, page-by-page plan review and the administrative backlog that slows permit issuance.

Pickle builds a personal memory layer for AI that continuously captures a user's real-world context, preferences, and intent so that AI assistants can act proactively on their behalf. The system combines a web-based app (Chat) and proprietary AR glasses (Pickle 1) to collect context across both digital and physical environments. It replaces the manual effort of re-explaining personal background, preferences, and goals to AI tools every time they are used.

Pinch builds an AI dubbing and live translation API for developers. The dubbing product translates video and audio into 10 languages while preserving the original speaker's voice, tone, and background audio. The live translation product converts speech in real time for meetings, webinars, and browser audio, replacing manual localization workflows and human interpreters.

Praxim builds an AI-powered Word editor that executes document-wide edits on command, replacing manual find-and-revise work. Writers and knowledge workers describe the changes they want in plain language, and Praxim applies them across the full document while respecting formatting rules, linked files, web context, and the user's personal preferences.

Proception Inc builds advanced hands for humanoid robots, targeting the dexterity gap that prevents robots from performing fine motor tasks such as threading a needle. The hardware combines robotics and AI to give humanoid platforms human-level manipulation capability. It replaces the need for custom, task-specific end effectors or human labor on precision assembly work.

Promptless keeps customer-facing documentation current for fast-growing startups and Fortune 500 enterprises. It monitors product changes, support tickets, and internal communications, then generates draft doc updates that technical writers review and publish. This replaces the manual, cross-team effort of tracking feature releases and editing docs by hand.

Quantstruct builds an AI documentation engineering platform for teams that maintain API and SDK references, implementation guides, and changelogs. It deploys multiple AI agents that test existing docs against a live product, identify gaps triggered by new releases or unanswered customer queries, and then generate updated documentation complete with embedded screenshots and video recordings by operating directly inside a web browser. The product replaces the manual work technical writers, professional services teams, and product teams do to keep documentation current after every product change.

Rally builds a financial operating system for vehicle fleet operators, targeting European businesses such as delivery services, construction crews, and field service companies. It consolidates fleet spending, including fuel, charging, and maintenance, onto a single platform and automates bookkeeping, credit, and driver payroll. The product replaces fragmented legacy payment networks and manual back-office workflows that fleet managers otherwise handle across multiple systems.

Rebolt builds an agentic operating system for commercial aircraft maintenance checks. It connects ERPs, databases, spreadsheets, and third-party systems across planning, engineering, production, materials, and billing, replacing the manual coordination that runs across those tools today. Maintenance teams use it to triage findings, draft repair instructions, schedule hangar capacity, manage task cards, and capture additional work for billing, all inside a single AI-driven workflow.

Red Barn Robotics builds an autonomous weeding robot, called the Field Hand, that removes intra-row weeds on farms without human intervention. The robot targets the hand-weeding work that farm labor crews currently perform, a job made increasingly costly by chronic agricultural labor shortages. It navigates fields independently, uses mechanical cultivation to protect soil health, and is designed to keep fields weed-free across a full growing season.

Reditus Space builds reusable satellites that carry customer payloads to orbit and return them intact. Pharmaceutical, materials, semiconductor, and aerospace defense customers use the platform to run microgravity manufacturing, hypersonic reentry testing, and spaceflight qualification work that currently requires either disposable hardware or no orbital option at all. The company handles the full mission stack, from launch integration through on-orbit operations to controlled reentry and physical payload return.

ReJot builds Fragno, an open-source framework that lets API companies ship complete, ready-to-use integrations instead of bare endpoints and documentation. Backend logic, data storage patterns, and frontend utilities are bundled together into reusable blocks that customer developers plug in and configure rather than build from scratch. The product targets API providers who want to cut the integration work their customers must do themselves.

Retrofit builds an AI-powered vintage marketplace that curates secondhand clothing for individual shoppers. It analyzes social media trends, resale data, and personal style preferences to surface vintage pieces matched to a buyer's taste, size, and budget. The product replaces manual browsing across thrift stores and resale platforms with a personalized, continuously updated selection.

Revise Robotics builds AI-powered robotic systems that automate laptop refurbishment for e-waste recycling facilities. The system tests devices, wipes hard drives, photographs them, and lists them on resale marketplaces, handling hundreds of units per day across any model, manufacturer, or operating system. It replaces the manual labor recycling facilities currently spend sorting, evaluating, and preparing discarded consumer electronics for resale.

Riviera builds AI voice agents for hotels, handling the inbound phone work that front-desk staff currently manage manually. Its first agent takes calls around the clock, answers guest questions, processes in-room dining orders, and books reservations in over ten languages, integrating with existing reservation, loyalty, and operations software. Hotels replace overwhelmed phone lines and after-hours voicemail gaps without adding headcount.

Roark builds a QA and evaluation platform for teams shipping voice AI agents. It scores every production call across hundreds of audio-native metrics, runs end-to-end simulations against configurable caller personas before launch, and automatically turns failed calls into repeatable regression tests. It replaces manual call review and ad-hoc testing scripts that miss edge cases and slow down deployment.

Rocketable acquires small, profitable software businesses and replaces their human workforces with coordinated AI agents, targeting B2B software products. The company builds the orchestration tools needed to direct agents across every business function, from support and marketing to resource allocation, eliminating the need for human employees entirely. The work it replaces is the full operating staff of a software company.

SalesPatriot builds a back-office operating system for defense manufacturers, resellers, and distributors competing for U.S. government contracts. It automatically surfaces relevant Pentagon solicitations and subcontractor RFQs, digitizes the bid preparation process, and connects internal departments that previously coordinated on paper. The product replaces the manual, repetitive work of hunting for contract opportunities, processing RFQ documents, and moving data between disconnected tools.

SAMMY Labs builds a knowledge layer for legal and compliance teams at large enterprises, connecting regulatory and policy updates to the internal processes they affect. When a compliance rule changes, SAMMY identifies which processes are impacted and rewrites them automatically, replacing the manual work of tracking, interpreting, and propagating updates across an organization. The product also serves as a knowledge substrate for AI systems, giving them accurate, structured context instead of raw document search.

Scout builds an AI-powered student information system for K-12 schools and districts, replacing legacy SIS platforms that require layers of add-on tools to cover gaps. It handles enrollment, attendance, scheduling, and compliance reporting through AI agents, cutting tasks like student onboarding and audit preparation from weeks of manual work to minutes. The product targets school administrators who currently piece together operational workflows across disconnected systems.

Sennu AI builds an AI code review tool for Salesforce development teams. It analyzes every pull request with full awareness of the existing Salesforce codebase, performing the review work that a senior Salesforce architect would otherwise do manually. The product targets engineering teams building on Salesforce who need consistent, context-aware code review without adding senior headcount.

Solidroad builds AI agents that train and evaluate customer experience reps. Before reps go live, the platform runs them through simulated conversations modeled on real customers, then scores their performance against custom rubrics covering communication, empathy, and process adherence. This replaces manual coaching, shadow sessions, and ad-hoc QA review at contact centers and BPOs.

Splash Inc. builds autonomous surface vessels for national security missions, including coastal patrols, harbor security, and protection of critical maritime infrastructure such as oil rigs and shipping terminals. The vessels operate without user input, carry variable payloads, and reach speeds of 50 knots with a 350-nautical-mile range. Splash targets the work currently done by crewed naval patrol craft, replacing it with scalable, mass-manufactured drone boats.

Spott builds an AI-native applicant tracking and CRM platform for recruiting and staffing firms. It replaces fragmented, manually intensive tools like Bullhorn, Loxo, and Vincere by combining ATS, CRM, and AI agents across the full recruiting workflow in a single system. Recruiters get semantic search, AI-driven candidate matching, multi-channel outreach, call intelligence, and candidate presentation tools without bolting on separate add-ons.

Stamp builds an AI-powered email and calendar assistant for individual professionals. It drafts replies in the user's voice, prioritizes and categorizes incoming mail, tracks action items, and schedules calendar events, replacing the manual triage and composition work people do themselves today. The product learns each user's preferences through stored memories, contact history, and custom workflow rules.

Steinmetz builds high-power-density motor controllers and traction inverters for electric vehicles and adjacent markets including robotics, aerospace, marine, and heavy equipment. Its 0-Series controllers deliver over 100 kW/L, outperforming comparable units from Tesla, Cascadia, and BMW, while matching them on cost. The product replaces incumbent inverters that force EV manufacturers to trade off power density against price or size.

Stillwind builds AI tools for electrical engineers. Its first product, Stillwind Search, lets engineers find electronic components by typing natural-language queries, which the engine converts into precise specifications matched against a database of millions of parts. A second product corrects PCB component libraries at scale, processing footprints directly from datasheets and manufacturer drawings to replace the manual, error-prone work that causes production failures.

Strike builds prediction markets for businesses, letting organizations tap collective intelligence to forecast outcomes on decisions and events. The product replaces traditional top-down forecasting and internal survey processes with a market-based mechanism where participants trade on expected results.

SubImage builds a security graph platform that maps cloud and on-premises infrastructure, showing security teams exactly what assets exist, how they connect, and where exposure and misconfiguration risks lie. It replaces manual spreadsheet tracking and blind-spot-prone point tools by providing continuous, graph-powered visibility across cloud providers, identity systems, and internal tooling. SubImage is the commercial, managed offering built on the Cartography open-source project, created by Cartography's original authors.

Sublingual builds a desktop productivity tracker for software teams. It keeps teams focused on shipping by monitoring daily progress, replacing informal check-ins and manual status tracking. The tool targets developer teams that need lightweight, continuous accountability without added process overhead.

Subtrace builds a zero-code network tracing tool for engineering teams running distributed infrastructure. It captures every inbound and outbound HTTP request in real time by attaching to any process at startup, using BPF under the hood, with no root privileges and no code changes required. It replaces the manual work of instrumenting services or piecing together logs to understand network-level behavior across an entire stack.

superglue builds AI agents that handle enterprise system integrations, replacing the brittle SQL scripts, cron jobs, and custom connectors that engineers and implementation teams maintain by hand. Its agents map and migrate data, configure ERPs like NetSuite, SAP, and Sage Intacct, and keep systems in sync after go-live. Enterprise software and operations teams use it to cut implementation time and reduce the ongoing cost of connecting ERP, CRM, and AI platforms.

Swerve builds an AI-driven interactive fiction app where users shape stories through their own words and choices, replacing static choose-your-own-adventure formats and passive reading. It targets adult entertainment and fiction consumers who want narrative experiences that respond dynamically to their input rather than following a fixed script.

TamLabs builds an AI plugin for Microsoft Word that lets professionals edit documents through natural-language commands. It applies changes instantly with version-control-style previews, resolves tracked comments automatically, and preserves existing formatting. The product targets knowledge workers in finance and legal who spend time manually redlining, reformatting, and implementing reviewer comments in Word.

Tejas AI builds a risk decisioning platform for banks, letting credit teams develop and evaluate credit rules faster and with greater confidence. It targets the manual, judgment-heavy process of setting and testing credit policy, replacing slow iteration cycles with a structured decision-making workflow.

TensorPool builds a GPU orchestration platform for machine learning engineers who need to train models without managing cloud infrastructure. Developers describe their training job via a CLI, and TensorPool handles GPU cluster provisioning and execution, replacing the manual setup typically required on major cloud providers.

Tergle builds AI agents that automate audit workflows for B2B companies, replacing the manual, repetitive work traditionally done by human auditors. Clients go through a tailored onboarding process where Tergle's engineers and domain-experienced auditors configure an AI solution to match their specific audit process. The product targets organizations that run recurring audits and want to eliminate the hands-on labor those cycles require.

ThirdLayer builds Dex, a browser-native AI copilot for knowledge workers and operators handling complex, multi-tool workflows. Dex maintains a live operational graph across messages, meetings, deals, and tasks, then uses that accumulated context to drive autonomous agents, fill forms, and execute work directly in the browser. It replaces the manual effort of context-switching, status tracking, and repetitive handoffs across tools like Gmail, Slack, and CRMs.

Toothy AI builds AI agents for dental practices that handle insurance verification, billing, and claims processing, including outbound insurance calls via voice AI. Dental clinics typically assign staff hundreds of hours a month to these administrative tasks; Toothy AI's agents automate the entire workflow end to end, reducing manual labor and accelerating reimbursements.

Trace builds voice AI agents that handle customer support calls end-to-end for banks, fintechs, and insurers. The agents connect directly to core banking systems and external APIs to complete tasks, such as card replacements, transaction lookups, fraud complaints, and account lockouts, without handing the caller off to a human or a self-service flow. This replaces the live agents and IVR menus that financial institutions currently rely on for high-volume, repetitive inbound call types.

TrainLoop builds specialized post-training models for enterprises in pharma, biotech, logistics, and banking, focusing on long-horizon tasks that require deep domain expertise. Working alongside human experts, TrainLoop distills their judgment into proprietary models through reinforcement learning and post-training research. This replaces the effort of fine-tuning general-purpose LLMs in-house or relying on off-the-shelf models that underperform on complex, domain-specific workflows.

Trata builds a research platform for buy-side analysts at hedge funds, giving them access to a curated library of transcribed conversations between anonymous portfolio managers and analysts who hold major positions in the stocks they cover. It replaces the manual, fragmented process of sourcing peer perspectives on public equities with a searchable, continuously growing archive of candid fund-to-fund dialogue. AI agents conduct and transcribe these conversations at scale, turning unstructured buyside debate into structured, distributable research.

Trava builds AI agents that audit customs entries and automate product classification for US importers. It connects to a company's PLM or ERP, reviews every entry for duty and classification errors, and catches overpayments before they compound. The work it replaces is the manual compliance review that customs brokers perform, at a pace and scale brokers can no longer sustain.

Trim builds a foundation model for physics that simulates real-world physical systems evolving over time. Where traditional simulators scale exponentially or polynomially with problem size, Trim's Transformer architecture uses linear-attention to scale linearly in both dimensions and grid size. This makes latency-sensitive tasks such as autonomous vehicle path planning orders of magnitude faster, and previously infeasible tasks such as gravitational wave detection computationally tractable.

Truffle AI builds infrastructure that lets developers embed AI agents into software applications via simple API calls. It replaces the need to architect and host custom agent backends from scratch. Developers use it to wire purpose-built agents, such as customer support or analytics, directly into products like WhatsApp and Slack.

Tweeks.io builds an AI recruiter that conducts voice-driven technical and non-technical candidate screenings for hiring teams. It evaluates candidates simultaneously at scale, scoring each one against a role-specific rubric and delivering structured results to recruiters within minutes. The product replaces manual phone screens and early-round interviews, letting hiring teams skip directly to deciding which candidates move to an onsite round.

Uncommon Therapeutics develops therapeutics targeting the root cause of rare genetic diseases, starting with Rett Syndrome, a neurodevelopmental condition affecting one in ten thousand girls. The company applies a polytherapy strategy, validating multiple drug modalities in disease-relevant models to advance a pipeline that individually improves patient outcomes and in combination aims to deliver a cure. This work replaces the standard single-target approach with a coordinated multi-drug design built around deep, firsthand disease expertise.

Vantel builds an AI-native operating system for commercial insurance brokerages, centralizing the workflows those firms run across documents, client records, and policy data. It replaces the patchwork of manual processes and disconnected tools brokers use to manage submissions, renewals, and client communication.

Verbiflow builds Mochi, an AI sales coworker that lives inside Slack and helps go-to-market teams prepare for calls, follow up on deals, and build pipeline. Mochi pulls context from a team's CRM, conversations, and documents to surface account history and open questions on demand. It replaces the manual work of digging through Salesforce, HubSpot, emails, and call notes before every customer interaction.

Vetnio builds an AI copilot for veterinarians that automates the administrative work consuming a large share of their clinical day. The platform transcribes and drafts medical notes in each vet's own style, provides AI-backed diagnostic support, and handles client calls and messages. It replaces manual note-taking and routine communication tasks so veterinarians can spend more time on patient care.

Vocality Health builds a voice AI system that translates medical conversations in real time, letting doctors speak with non-English-speaking patients directly during in-person and video visits. It replaces human medical interpreters, including the call-center-based remote services hospitals currently rely on, which are costly and frequently leave limited-English-proficient patients underserved.

Waypoint Transit builds AI-powered city planning software for municipal governments and planning firms. It automates the production of civil infrastructure studies, including corridor studies, safety audits, development reviews, and GIS analyses, replacing work that cities currently contract out to expensive consultants. Planners connect their existing data sources and generate reports in a fraction of the time and cost of traditional methods.

Weave builds an engineering intelligence platform that measures how software teams and AI coding agents deliver work. It tracks pull requests, deploys, AI tool usage, token spend, and developer feedback to surface bottlenecks from prompt to production. Engineering and platform teams use it to replace manual guesswork about where delivery slows down and whether AI adoption is improving output or burning budget.

Wildcard builds an optimization platform for e-commerce and retail brands that want their products recommended by AI shopping assistants such as ChatGPT, Gemini, and Perplexity. Brands connect their product catalog, then track SKU-level rankings across real buyer queries, identify why specific products are surfaced or skipped, and get a prioritized list of attribute fixes to improve both discovery and conversion. The platform replaces manual guesswork about AI recommendation logic with structured, prompt-level visibility and a concrete action roadmap.

Woz builds an AI platform that lets non-technical entrepreneurs design, launch, and grow mobile apps without writing code or hiring engineers. Its agent-based system handles product development, app store submission, and business scaling end to end, replacing the need for a technical co-founder or in-house engineering team. On the developer tooling side, Woz also ships WozCode, a Claude Code plugin that cuts token costs and speeds up coding sessions for software engineers running parallel AI-assisted projects.

YouShift builds AI-powered scheduling software for hospital workforce teams, replacing manual spreadsheets and legacy scheduling tools. AI agents handle day-to-day coordination, carrying a single record from scheduling rules through worked hours to gross pay. Clinicians, managers, and executives each get a tailored view, letting staff set availability by voice, swap shifts, and check pay, while schedulers are freed from routine coordination work.

ZeroEntropy builds specialized retrieval models, including rerankers and embedding models, for developers shipping AI products. Its API sits between a knowledge base and an LLM, replacing standard semantic search with higher-accuracy retrieval to cut hallucinations in Q&A bots and AI agents. Target customers span legal, finance, healthcare, manufacturing, and e-commerce teams building on document-heavy pipelines. ZeroEntropy has since joined Notion.