YC Fall 2025 (F25) companies

Absurd builds a production canvas for creating AI-generated brand and performance video ads. Marketing teams at B2B software companies use it to go from concept to finished video in 72 hours, replacing the longer cycles of traditional video production agencies. A multi-agent orchestration layer handles the creative pipeline while keeping filmmakers, clients, editors, and legal stakeholders in sync on a single shared workspace.

AirCaps builds smart glasses that display real-time captions, translations across 60-plus languages, and AI-generated meeting notes directly in the wearer's field of view during face-to-face conversations. The product serves deaf and hard-of-hearing users, international travelers, and professionals such as sales reps, doctors, and lawyers. It replaces manual note-taking, human interpreters, and the post-meeting summarization tools that only work after a conversation ends.

Aleph Lab builds an AI language tutor for children that lives inside Minecraft, where kids practice speaking a new language by conversing with an AI companion named Annie during normal gameplay. It targets parents of school-age children who want spoken language practice but lack access to native-speaking teachers or tutors. The product replaces traditional language classes and passive screen time with active, conversation-driven practice inside a game children already play.

Allus AI builds vision foundation models for manufacturing plants. The models watch production in real time, interpret what is happening on the floor in operational terms, and surface intelligence on safety, quality, delivery, and cost. This replaces manual observation and disconnected monitoring tools that cannot reason about what they see.

Amera automates claims processing for self-funded health insurance payers and plan administrators. It ingests claims submitted in any format, including PDFs, paper bills, receipts, EDI files, and non-standard submissions, converts them into structured data, and routes that data into existing payer systems. This replaces manual data entry and exception handling by claims operations teams.

AnswerThis builds an AI research workspace for academics, doctors, and scientific teams who need to conduct literature reviews. It searches and analyzes over 300 million papers, clinical trials, and PDFs to surface relevant sources, map connections between ideas, and generate cited answers to complex research questions. This replaces weeks of manual screening, note-taking, and synthesis across fragmented tools with a single end-to-end workflow that also produces compliant systematic literature reviews.

Anto Biosciences builds foundation models for microbial communities, making the gut microbiome computationally tractable for drug development. Its Darwin model series predicts how gut bacteria alter drug toxicity and efficacy across diverse patient populations, then optimizes molecules to improve outcomes. This replaces the current practice of treating microbiome variation as noise during drug development, addressing a root cause of clinical drug failures.

Arcten builds long-horizon autonomous agents for research and coding tasks. Software teams use these agents to handle complex, multi-step work that would otherwise require sustained human attention across research and development workflows. The YC program description points to embedded AI copilots that let end users drive those workflows through natural language, replacing manual coordination inside existing products.

Articulate builds an AI platform that reads 2D PDF construction plan sets and automatically flags coordination conflicts, missing information, and cross-discipline mismatches before construction begins. General contractors, architects, and engineers use it to replace slow manual drawing reviews and multi-day page-turn meetings. The platform drafts RFIs from detected issues and pushes them directly into Procore or Autodesk, cutting rework and keeping projects on schedule.

Aside builds a real-time answer copilot for enterprise tech sales reps. During live calls, it listens and pulls relevant answers from internal docs, past winning calls, and connected tools, so reps can respond to questions and follow the team's sales playbook without pausing to search. It replaces post-call coaching and manual knowledge retrieval with in-call guidance.

Aspect builds an intelligent media storage and collaboration platform for teams working with large video and multimodal datasets. It replaces manual file shuttling, scattered review tools, and ad-hoc quality checks by centralizing upload, local streaming, version review, segmentation, and AI-driven schema extraction in one platform. Data teams at AI labs use it to wrangle dense multimodal datasets the way text-focused teams use document and search tools for large text corpora.

AtlasGrid builds a visual analytics layer for mobile app product teams, mapping UI screens into an interactive graph and overlaying usage data directly on each screen. AI agents continuously scan thousands of mobile apps to surface design patterns, growth opportunities, and concrete product improvements. The tool replaces manual competitive benchmarking and the disconnected workflow of correlating analytics dashboards to specific app screens.

AutoAce builds an AI-native operating system for car dealerships, starting with AI service advisors that answer inbound calls and schedule appointments. The platform is designed to unify service, sales, parts, finance, and insurance into a single system, replacing the fragmented, reactive workflows dealership staff handle today.

Automax.ai builds an AI-powered residential property appraisal platform for mortgage lenders and licensed appraisers. A mobile app uses LiDAR and computer vision to capture floor plans, photos, and condition data in a single walkthrough, then AI agents pull comparable sales and generate a complete, Fannie Mae and Freddie Mac compliant appraisal report in under 20 minutes. The platform replaces a process that typically takes five to ten days of queuing, scheduling, and manual report writing.

Autonomous Technologies Group builds an AI-native wealth management platform for individual investors, replacing both human financial advisors and cookie-cutter robo-advisors. It monitors accounts across 401(k)s, taxable brokerages, equity holdings, real estate, and cash flow, then executes sophisticated strategies, such as tax-loss harvesting, portfolio rebalancing, QSBS tracking, and borrowing against assets, that traditional advisors reserve for ultra-high-net-worth clients. The product handles the ongoing coordination and decision-making that investors currently do manually or pay a human advisor to do.

Axial Composites Industries builds a non-planar manufacturing cell that produces carbon fiber parts from thermoplastic composites in days rather than weeks. It targets aerospace and robotics manufacturers who need structural composite parts, replacing traditional layup and curing workflows that are too slow for rapid iteration. The system exploits fiber anisotropy to produce parts that are both stronger and lighter than conventional composite fabrication methods yield.

Bear builds a marketing stack for companies that want to capture and convert high-intent visitors arriving from AI platforms such as ChatGPT and Perplexity. It covers the full funnel from the moment an AI agent surfaces a brand to the point of conversion. The product replaces ad-hoc, piecemeal attempts to track and act on AI-referred traffic.

Bluma automates short-form video ad production for brands, performance marketers, agencies, and creators. It generates UGC-style videos, template-driven social content, and paid social creatives from a prompt, template, or source reference, replacing the scripting, production, and editing work that growth and creative teams do manually. The output is optimized for TikTok, Instagram Reels, and YouTube Shorts at high volume.

Bolna AI builds a voice AI orchestration platform for Indian enterprises that need to run high-volume inbound and outbound calls across sales, support, collections, and recruitment. Companies feed in their transcripts and FAQs to create and deploy voice agents that handle calls in Hindi, Tamil, Telugu, Hinglish, and other Indian languages, replacing manual calling teams and third-party call center operations.

Boom AI builds an AI agent for e-commerce brands that handles the full customer lifecycle across channels including WhatsApp, Instagram, TikTok, Facebook, and web chat. The agent answers product questions, recovers abandoned carts, supports orders, and runs outbound marketing campaigns, replacing the work of a dedicated sales and support team. Brands connect it directly to their store without manual flow-building.

Bravi builds an AI operating system for home-services businesses, starting with fenestration trades such as shutters, windows, and garage doors. On the customer-facing side, AI agents handle inbound calls, chats, and follow-ups so installers stop losing leads to missed contact. On the field side, a copilot connected to product catalogues, ERP data, and technical documentation lets manufacturers' sales reps answer spec questions, configure products, and close quotes on the spot, replacing back-and-forth with technical support teams.

Brickanta builds an agentic AI platform for construction estimation and procurement, serving mid-market subcontractors, large general contractors, and real estate developers. Hundreds of AI agents, trained on standardized building protocols, analyze project details and estimates to automate bid preparation, flag risks, and consolidate company knowledge. The product replaces manual document work and error-prone estimation workflows that slow project delivery and inflate costs.

Brickwise builds an AI property manager for landlords and estate agents, handling maintenance requests, tenant and guest communication, contractor follow-up, leasing, and compliance around the clock. Property managers shift from coordinating work to simply approving it. The product serves operators across asset classes including multifamily, single-family rentals, student housing, and short-term rentals.

Caddy builds a proactive personal AI assistant that works across email, calendar, Slack, and other everyday tools to handle tasks and communications on the user's behalf. It monitors schedules, messages, and to-dos, then acts without waiting to be asked, drafting replies, booking appointments, and updating documents in the user's own voice. It replaces the manual coordination and context-switching that fills a knowledge worker's day.

Casey sells commercial insurance to small business owners in industries like trucking, contracting, and restaurants. A specialist gathers the business details, shops more than 100 carriers, and delivers bound coverage with ongoing certificate and claims handling. Casey replaces the back-and-forth with traditional brokers who are slow to respond and rarely shop the market at renewal.

Castari builds deployment and runtime infrastructure for AI agents, letting developers ship agents to isolated, autoscaling cloud sandboxes in a single command. It handles MCP tool connectors, environment isolation, logging, and observability, replacing weeks of custom infrastructure work. The target customer is a developer team moving an AI agent from prototype to production.

Clad Labs builds an AI support agent that triages, investigates, and resolves customer issues without requiring engineering involvement. It ingests threads from Slack, Discord, Teams, and email, then traces problems to root cause by pulling analytics, backend data, and incident history. The agent replaces the manual work of context-gathering, handoff coordination, and ticket-writing that support and engineering teams otherwise do by hand.

Claybird produces launch films, feature announcement videos, fundraise films, and event cuts for technology companies. Each project is scoped to a specific moment in a company's story, replacing the need for a traditional video production agency. Work is delivered in seven days through a shared Slack channel with the client.

Clicks deploys AI agents that handle repetitive back-office tasks for B2B companies, starting with recruiting and executive search firms. A forward-deployed engineering team learns each client's existing workflows through video calls or documentation, then automates work like candidate data entry, CRM updates, and assessment reports without requiring IT integrations. Agents run around the clock, replacing the manual administrative labor that staff would otherwise do by hand.

Codyco builds an AI phone reservation system for hotel groups. It answers incoming and missed calls around the clock, handles bookings, payments, and guest questions, and writes confirmed reservations directly into the hotel's property management system. The product replaces the manual work of front-office staff and central reservation teams fielding routine calls.

ComplyDo builds a compliance automation platform for enterprise GRC teams, replacing manual work across regulatory mapping, evidence collection, and gap analysis. It ingests regulations, standards, contracts, and internal documentation, then constructs a structured graph that links every requirement to relevant controls and evidence. GRC teams use it to track coverage, close gaps, fill questionnaires, monitor regulatory changes, and manage compliance across multiple products and legal entities.

Compyle builds an AI platform for environmental due diligence firms, specifically for Phase I Environmental Site Assessments. It automates desk work that Environmental Professionals currently do by hand: collecting historical aerial photos, Sanborn fire insurance maps, city directories, and regulatory database records, then surfacing findings and drafting reports in the firm's own template. This replaces manual data ordering from vendors like EDR or ERIS and the multi-day wait that comes with it.

Cortex AI collects and supplies large-scale, real-world robotics training data to frontier AI labs building foundation models and general-purpose robots. Its datasets cover egocentric human video annotated with hand and body pose, depth, and subtask labels, plus robot trajectories gathered from manipulators and humanoids in live industrial settings. The work it replaces is the slow, expensive, and narrow data collection labs would otherwise run themselves.

Cranston AI builds accounting software for CPA firms and SMBs that prepares tax returns, reconciles books, and handles compliance work inside the tools firms already use, including Drake, QuickBooks, and CCH Axcess. The product reads client documents, classifies transactions, drafts journal entries, and writes client follow-up emails, replacing the manual preparation work done by accounting staff. Reviewers approve everything before it posts, files, or reaches a client.

Crunched builds an AI analyst embedded directly in Excel for finance and consulting professionals in consulting, investment banking, and private equity. It audits entire workbooks for errors, constructs financial models from scratch, and pulls structured data from information memoranda into template models, replacing the manual grunt work that junior analysts and senior practitioners spend hours on. The product operates inside Excel so teams review and revise output through their existing workflow.

Dari builds an API for automating browser-based workflows, targeting developers and businesses that need reliable web agents. Teams define workflows in natural language, and Dari executes them through deterministic playback, falling back to large language models only when needed. This replaces brittle scraping scripts, manual web tasks, and custom browser automation code.

Deeptrace builds AI agents that handle on-call investigation and resolution for engineering teams. When a production alert fires, Deeptrace automatically pulls in logs, traces, metrics, and code, identifies the root cause, and delivers evidence-backed findings to Slack, replacing the manual triage and context-switching that pulls engineers away from building. It targets engineering and reliability teams at companies that run production systems and cannot afford extended downtime.

Denki automates internal financial auditing for companies subject to SOX and SEC regulations, handling the planning, control testing, evidence collection, and documentation that audit teams do manually today. It operates inside existing tools such as AuditBoard, Workiva, and ERPs, and produces workpapers with a full audit trace at each step. The product replaces the bulk of labor-intensive audit execution work that would otherwise fall to internal audit staff or outside firms.

Digipals builds an AI-native social platform that consolidates fragmented friend communication into a single group-chat environment. Ambient AI agents pull context from users' location, calendar, and photos to take actions and automate tasks directly inside those chats. The product replaces the patchwork of iMessage, WhatsApp, Partiful, Venmo, and Find My that friend groups currently juggle across separate apps.

DiligenceSquared automates commercial due diligence for private equity funds, replacing the weeks-long, six-figure engagements that traditional strategy consultancies deliver as static, unauditable slide decks. Its AI agents conduct expert interviews, gather market data, and synthesize findings with every claim linked to a source transcript, while senior consultants with prior MBB experience direct and review the work.

Dome builds a unified API layer for prediction markets, targeting developers who want to build on platforms like Kalshi and Polymarket. It provides access to live and historical market data, including granular datasets those platforms do not expose natively, through a single interface. Developers use it to trade programmatically, embed market data into products, or deploy strategies across multiple platforms, replacing the need to integrate each exchange separately.

Efference builds perception hardware and software for robots, replacing conventional stereo cameras with a software-defined approach to 3D sensing modeled on human vision. Its devices combine robotics-grade sensors, hardware-synchronized compute, and an integrated stack covering data collection, teleoperation, and inference. The product targets robotics teams that need high-fidelity spatial perception at lower cost and complexity than existing depth hardware.

Everest builds AI employees for managed service providers, the outsourced IT firms that handle support tickets on behalf of their business clients. The product automates ticket resolution work that MSP support engineers currently handle manually, compressing the labor-intensive steps that keep margins thin in the outsourced IT industry.

Exonic builds an open, crowdsourced drug discovery platform that lets researchers worldwide combine biological AI models, design workflows, and run computational experiments without writing code. It targets the closed, institution-siloed structure of traditional drug discovery, replacing proprietary internal pipelines with a meritocratic tournament model where anyone can contribute. The platform includes a wet lab validation layer, as demonstrated by a benchmark result in synthetic enhancer design for liver cancer.

Expected Parrot builds an AI simulation platform that lets companies run interviews and surveys with LLM-powered agent personas modeled on their customers. Teams use it to test pricing, product, marketing, and messaging scenarios at scale before committing to real research. The platform replaces or precedes traditional survey work, and supports validation by sending the same surveys to human respondents within the same interface.

Fastshot builds a no-code AI platform that turns a plain-language description into a native iOS and Android app, ready to publish on the App Store and Google Play. It targets non-technical founders and entrepreneurs who would otherwise hire developers or spend months in a build cycle. Fastshot handles design, branding, assets, backend, authentication, and store metadata end-to-end.

Fernstone builds an AI-powered commercial insurance brokerage for businesses with complex risk profiles, starting with construction and security companies. It handles the full brokerage workflow: obtaining coverage, issuing certificates of insurance, managing renewals, and monitoring policy gaps. The product replaces traditional commercial insurance brokers, combining a dedicated risk advisor for each client with a software platform for coverage management.

Fifth Door lets anyone build, share, and play games using AI, without requiring design or coding skills. The platform targets everyday users who want to turn creative ideas into playable games and share them with friends. It replaces the traditional barrier of needing a development team or technical expertise to bring a game concept to life.

Fixpoint builds a marketplace that sources, vets, and deploys human annotators and AI trainers onto data labeling and model evaluation projects. It serves AI teams that previously staffed these roles through slow, manual recruiting. The platform automates annotator matching, qualification checks, and payroll, replacing a fragmented staffing process with a managed network of domain experts.

Flick builds an AI filmmaking platform for creators who want to produce structured, cinematic short films rather than disposable social clips. The interface supports end-to-end film production, keeping characters consistent, storylines coherent, and the creative process unified in a single environment. It replaces the fragmented stack of disconnected AI generation tools that currently force filmmakers to stitch together outputs with no continuity or directorial control.

Forge Robotics builds modular, autonomous factories for metal fabrication, targeting manufacturers facing structural skilled-labor shortages. Each shipping-container-sized module uses sub-millimeter vision and onboard AI to plan, weld, scan, and verify parts end-to-end, replacing manual welding crews and rigid production lines. A manufacturer adds modules to expand capacity without adding headcount.

Freeport Markets builds a DeFi brokerage app that gives wealth-oriented retail investors access to pre-IPO companies, global equities, crypto, commodities, and structured credit through perpetual futures and onchain instruments. It targets investors locked out of private markets or foreign exchanges, replacing the need for multiple brokers, fund minimums, or accredited-investor gatekeepers. An integrated AI layer converts news sources into actionable trade ideas within the same app.

Goldbridge builds an integrated financial platform for real estate owners and operators, combining banking, spend management, and automation in a single system. It targets a segment that has historically stitched together generic banking tools with manual back-office work, and replaces that patchwork with purpose-built financial infrastructure for the real estate vertical.

Hypercubic builds an agentic AI platform that helps large enterprises maintain and modernize COBOL codebases and mainframe systems. Its tools convert opaque legacy code into searchable documentation, capture expert debugging workflows as digital twins, and provide a natural-language terminal for mainframe operations. The platform replaces manual archaeology of decades-old systems and fills the gap left as the original developers who built and understood them retire.

Hyperspell builds a persistent memory layer for AI agents, giving them access to a continuously updated, permission-aware knowledge graph built from a company's existing tools: Slack, Gmail, Notion, Linear, HubSpot, and more. Developers connect Hyperspell via API or SDK so their agents can recall past decisions, understand relationships between people and projects, and act on current context rather than isolated text fragments. It replaces the manual work of wiring agents to disparate data sources and the lost institutional knowledge those agents would otherwise lack.

Icarus builds autonomous, solar-powered aircraft that operate at 60,000 feet for weeks at a time, delivering persistent surveillance and communications coverage to defense customers. Flying above weather in the stratosphere, each platform is priced at roughly $100K and produced at scale in the thousands. The aircraft replace the role traditionally filled by expensive crewed reconnaissance planes and satellite assets.

Imagine AI builds AI-powered content agents for B2B founders and executives, turning their expertise, voice, and client insights into a continuous pipeline of LinkedIn posts, strategic comments, and long-form content. The product replaces the hours founders spend writing and distributing content themselves, while targeting specific ideal customer profiles using a database of over one million tracked professionals. The output is pipeline: booked meetings and inbound deals at named accounts.

Inspector builds an AI-native IDE for front-end engineers that links the browser directly to the codebase. Developers select any visual element in the browser, prompt a change, and coding agents receive the same runtime context a developer sees: rendered visuals, DOM elements, and console logs. This replaces the manual work of copying screenshots, switching tabs, and hunting through code to explain UI state to an AI tool.

Iris builds a personal AI assistant for calendar and email that learns how a user actually works: their reply style, priorities, and decision patterns. It monitors connected accounts, surfaces what needs attention, and pre-fills actions such as rescheduling meetings, drafting replies, and following up on threads, so the user can approve and hand off with a swipe. It replaces the manual triage and composition work people do across their inbox and calendar every day.

item builds an AI-native CRM that replaces traditional pipeline tools for sales, support, and account management teams. It pulls context from over 100 connected tools, maintains a unified customer record, and runs business processes autonomously through AI agents trained on plain-language process documents. The product displaces manual data entry, deal updates, lead research, and cross-app busywork that would otherwise fall to revenue-facing employees.

Jarmin deploys AI employees for technical and business roles, starting with machine learning engineering, for companies that need ongoing AI/ML work done without hiring full-time staff. It takes on full initiatives or individual tasks, training and fine-tuning models, building agentic systems, automating repetitive workflows, and running evaluations, replacing the day-to-day output of ML engineers, data scientists, and researchers. The AI employee learns a company's processes and standards, works inside existing tools and data systems, and communicates with the broader team like a human hire would.

JSX Tool builds a browser extension and in-browser IDE for React developers. It maps running UI elements back to their source components and files with a click, letting developers make precise visual requests to LLMs without switching to a separate editor or writing verbose code descriptions. It replaces the manual process of hunting down component locations and context-switching between the browser and a desktop IDE during React development.

Kaigo Health builds an AI-powered care management platform for Medicare outpatient programs. It automates post-discharge follow-ups, routine check-ins, and ongoing patient monitoring through phone calls, collecting voice biomarkers and remote monitoring data to support preventive care. The platform replaces the manual outreach and coordination work that care teams would otherwise handle person by person.

Kalpa Labs builds generalist audio models that understand and follow instructions, learn new tasks from context, and unify capabilities like speech-to-text, text-to-speech, and voice cloning into a single system. The first product is a conversational speech model that replicates not just speaker identity but delivery nuance, including pacing and emotional tone. It replaces the patchwork of single-purpose speech APIs that developers currently stitch together.

Karumi builds an AI agent that conducts live, interactive product demos over video calls for B2B software companies. It handles prospect and customer conversations across marketing, sales, onboarding, and enablement, replacing the work of sales engineers, customer success managers, and human demo specialists. The agent runs on demand, around the clock, in any language, so buyers can explore a product in depth without waiting for a human to be available.

Kestrel AI builds an AI agent platform for platform engineering teams, automating incident response, cloud provisioning, CI/CD pipelines, and developer requests. Engineers describe a workflow in plain English and Kestrel generates a deterministic, production-ready automation graph that executes across their stack, replacing manual triage, root-cause investigation, and remediation work that previously took hours.

Koyal builds an agentic AI filmmaking platform that converts scripts or audio into fully produced video stories, replacing manual animation, editing, and lip-sync work done by video production teams. Music labels, podcasts, and production houses use it to generate personalized, multilingual music videos and narrative content at scale. The platform covers script-to-audio-to-video in one pipeline, with state-of-the-art lip-sync, dynamic character movement, and multi-character consistency.

Lakonia builds an AI command-and-control platform for law enforcement and military commanders. It transcribes live radio communications, analyzes the content, and delivers synthesized operational intelligence to commanders, dispatchers, and field operators. The platform replaces manual radio monitoring and the cognitive work of piecing together fragmented voice traffic during active operations.

Lapis builds AI agents that handle the full paid advertising workflow for startups, from creating ad creative and launching campaigns across Google, Meta, LinkedIn, and ChatGPT, to generating matching landing pages and tracking conversions. The agents continuously optimize based on real traffic results, replacing the work of a dedicated growth marketing team. Startups use Lapis to acquire customers without hiring or managing the campaign operations themselves.

Leadbay qualifies B2B leads in markets where data and signals are scarce, including construction, hospitality, manufacturing, and healthcare. Sales teams use it to prospect accounts that conventional filter-based or signal-based tools cannot surface, replacing the manual research and judgment a top sales rep would otherwise apply. The platform learns from a team's won and lost deals, then feeds qualified, ranked accounts directly into AI agents running on Claude, ChatGPT, or Copilot.

Lemma builds production monitoring and continuous improvement infrastructure for AI agents. It audits every agent trace against the agent's own instructions, surfaces recurring failures, pinpoints the exact steps that cause them, and delivers fixes automatically via API or pull request. Teams replacing manual prompt tuning and reactive debugging with Lemma get agents that correct themselves in production rather than degrading silently.

Lexi builds an AI operating system for corporate law firms, handling drafting, document review, legal research, and matter management in one platform. It replaces the manual grunt work attorneys do across repetitive legal tasks: reviewing contracts, hunting for case law, and assembling case files from scattered emails and documents. Law firms use it to absorb more cases without growing headcount.

Lightberry builds conversational AI software that gives robots the ability to listen, speak, and act autonomously. Manufacturers and end users can program a robot by speaking to it directly, replacing traditional code-based configuration. The system runs on hardware from partners like Unitree and handles interaction and decision-making without human intervention.

Locus builds payment infrastructure for AI agents, letting developers and businesses give their agents access to paid APIs, data, and services with defined spending limits. Platforms connect once to a pool of nearly 4,000 tools through a single integration and contract, replacing the process of negotiating, contracting, and billing with each vendor individually. Every transaction is traceable and reconcilable with existing financial systems.

Logical builds a proactive desktop AI assistant that works across applications without requiring users to prompt it. It carries context between apps, drafts emails, writes documents, logs meetings, and surfaces action items automatically. The product replaces the manual work of switching tools, copying context, and typing detailed prompts into a chatbot.

LogosGuard builds AI risk management software for enterprises adopting AI tools, replacing manual policy reviews and ad-hoc vendor checks. It converts AI governance policies into automated controls, stress-tests AI products for compliance gaps, and continuously monitors for changes. A browser extension catches and redacts sensitive data, such as names, account numbers, and medical details, before prompts reach external AI tools, keeping all detection local to the device.

Lua Global Inc builds an AI agent operating system for mid-market businesses, letting engineering teams write agents in TypeScript or letting non-technical staff assemble them with a no-code builder on the same platform. The agents handle sales, customer support, and operations across channels such as chat, SMS, voice, and Slack, replacing work previously done by human teams or requiring bespoke enterprise AI infrastructure.

LunaBill builds AI voice agents that handle insurance claim follow-up for healthcare billing teams. The agents check payer portals, wait on hold, speak with payer representatives, draft and file appeals, and write every outcome back to the EHR, replacing the manual AR work that consumes the bulk of a billing team's day. Hospitals and revenue cycle management firms use LunaBill so their billers review results rather than chase claim status.

Lunavo builds an AI execution layer for freight forwarders and carriers, handling the back-office work that dispatchers, billing clerks, and tracking teams do today. It connects to a carrier's TMS, email, and portals, then autonomously handles order entry, shipment tracking, appointment coordination, billing preparation, and exception resolution. Work that previously required linear headcount growth scales without adding staff.

Mantle builds internal AI agents that automate back-office work for B2B companies. Founders and operators describe a task in a single prompt, and Mantle connects to tools like CRM, email, calendar, payments, and product analytics to run that work automatically. It replaces manual reporting, pre-meeting research, and routine email drafting that employees otherwise handle themselves.

MarkIt builds AI agents that automate trade compliance work for suppliers, customs brokers, and OEMs. Its product, Access, handles per-SKU regulatory review covering 21 CFR, MoCRA, and FTC claims, plus channel requirements for retailers such as Amazon and TikTok Shop, consolidating label artwork, formulas, ingredients, and claims into a single review. It replaces manual document review across Excel, PDFs, and broker portals, reducing classification errors and cutting time to market for international products entering the U.S.

Mayflower builds an AI plug-in that automates immigration screening and compliance for HR teams hiring international candidates. It parses candidate documents, classifies visa pathways, and produces attorney-ready forms, replacing work that previously fell to legal counsel or manual HR review. The platform integrates with existing HR systems so hiring teams surface only compliant candidates before legal review begins.

Metorial builds agentic infrastructure for AI-native companies, giving developers a serverless runtime to connect AI agents to external tools and data sources at scale. It handles the state management, access control, security, and tracing that teams would otherwise have to wire together themselves. The platform targets companies deploying AI agents across both customer-facing workflows and internal operations.

Miniswap builds a dedicated resale marketplace for tabletop wargaming miniatures, starting with Warhammer. Collectors and hobbyists buy and sell secondhand miniatures against a structured catalogue of over 20,000 distinct items, replacing the guesswork of general marketplaces where poor filters, free-form listings, and counterfeits make finding specific miniatures impractical.

Mixy builds a mobile app that lets anyone create song mashups by blending vocals and beats from separate tracks, no DJ skills required. It targets music fans who want to remix and share audio content but lack the technical knowledge to use professional mixing software. The app replaces both the learning curve of DJ tools and the manual work of audio editing.

Mod AI builds AI agents that handle accounts payable work for enterprise finance teams, replacing accounting clerks across the full payables workflow. The agents take on vendor communication, invoice coding, three-way matching, fraud detection, and exception handling, including edge cases like multi-currency invoices, partial deliveries, and split PDFs. The product targets the operational layer of corporate accounting that today depends on manual, clerk-level labor.

Moss builds a real-time semantic search runtime for voice agents, copilots, and chat interfaces, replacing cloud-based vector database retrieval pipelines that introduce lag. The engine runs natively in browsers, on edge devices, and on servers via a Rust and WebAssembly vector index, returning results in under 10 milliseconds without external network hops. Teams building conversational and multimodal AI products use Moss to eliminate the retrieval bottleneck that breaks the illusion of a responsive, intelligent agent.

MOVEdot builds an agentic engineering platform for automotive hardware development. Its AI agents run tasks, such as correlating test data against simulation models, extracting KPIs from databases, analyzing how environmental conditions affect performance, and generating reports, that engineers would otherwise spend hours or days completing manually. OEMs and tier-1 suppliers use MOVEdot to compress vehicle development cycles by orchestrating these agents across tools like Adams, VI-grade, MATLAB, and Canopy, inside their own cloud environments.

Multifactor builds zero-trust authentication, authorization, and auditing infrastructure for AI agents and multi-agent systems. It lets enterprises grant agents access to online accounts with fine-grained permissions, instant revocation, and cryptographically signed audit trails, replacing the ad-hoc credential sharing and password managers that leave agentic workflows exposed. The primary buyers are enterprise security teams, particularly CISOs, who need verifiable, auditable control over what AI agents can do inside connected accounts.

Narrative turns live and archived sports video into highlight packages, per-player data records, and frame-accurate event tags for broadcasters, leagues, teams, and high school programs. Its platform reads every play as it unfolds, finds viral moments, and sequences them into publishable clips while the game is still current. The work it replaces is manual video review, tagging, and highlight editing done by production and operations staff.

Nessie aggregates AI conversations and documents from tools like ChatGPT, Claude, and Cursor into a single, searchable knowledge base that individuals and teams own and build over time. It replaces the manual work of copying context between sessions and starting from scratch each time a new chat begins. Founders, researchers, writers, and their teammates can query accumulated history, surface past decisions, and share a living "brain" with collaborators instead of sending static documents.

Nexus builds AI agents for non-technical teams at large enterprises, turning documented processes into automated procedures that run without engineering support. Business teams configure and refine the agents themselves, replacing manual, expert-dependent workflows. Nexus engineers build the first version alongside the client team and continue shipping improvements tied to a specific business outcome.

Nivara builds an analytics layer for engineering leaders who need to understand how their teams are using AI coding tools. It tracks how humans and AI collaborate across the development workflow, surfaces bottlenecks in AI adoption, and measures the impact of that adoption on team output. It replaces ad hoc guesswork and manual observation that engineering managers currently rely on to evaluate AI tool effectiveness.

Nori builds an AI health coach for consumers who track their health with wearables, apps, and labs but struggle to act on the data. It pulls in data from devices like Oura, Whoop, Garmin, and Apple Health alongside nutrition logs and medical records, then surfaces a prioritized daily action plan with explanations. It replaces the work of manually cross-referencing fragmented health data and fills the gap left by expensive human health coaches.

Norra builds an asset management system for skilled nursing facilities, combining physical trackers with AI agents to give operators real-time visibility into the location, availability, and utilization of medical equipment such as beds, concentrators, and wheelchairs. Facilities today run these operations blind, relying on manual processes that result in significant equipment waste. Norra replaces that guesswork with continuous, connected asset intelligence across single sites and multi-facility networks.

Nucleo automates CT scan analysis for oncologists and radiologists, covering tumor lesion sizing, target versus non-target lesion classification per RECIST criteria, and body composition assessment for sarcopenia. The platform replaces manual segmentation and interpretation work that clinicians perform today. Hospitals including Stanford, Cedars-Sinai, UCI Health, and Weill Cornell use it to support tumor characterization and treatment decisions.

Null Labs builds simulation and synthetic data infrastructure for defense autonomy engineers developing perception systems on UAS, USV, and UGV programs. Its platform delivers physics-accurate sensor simulation, labeled synthetic imagery across operating conditions, and closed-loop software-in-the-loop testing, replacing slow real-world data collection and field testing cycles. Teams use it to cover edge cases and validate perception before hardware deployment across aerial ISR, maritime, ground, and counter-UAS domains.

OpenBuilder gives non-technical founders a no-code, AI-assisted platform to build and ship software without writing code. It replaces the stop-start cycle of conventional vibe-coding tools by offering unlimited build credits, active unblocking when users get stuck, and full access to the underlying codebase from the start. The target user takes an idea all the way to a production-ready product without handing off to engineers.

Openroll builds an AI-native platform that centralizes compensation, headcount planning, and budgeting for HR and finance teams. It connects to existing people and finance systems and automates the analysis, reviews, and approvals that previously required manual spreadsheet work. The product replaces fragmented tools and ad-hoc modeling by running compensation reviews, pay strategy tracking, and actuals-versus-budget analysis in one auditable system.

Parametric builds robots for frontline physical labor, starting with commercial laundry. Where conventional robots train on large generic datasets, Parametric's robots learn from each customer's own quality standards through an automated reward pipeline that pairs direct customer feedback with a judge model. The result is a robot that can meet the specific, idiosyncratic requirements of individual physical businesses rather than a one-size-fits-all average.

Parrot builds a Spanish-learning app that delivers short-form videos of native speakers, replacing grammar-drill and flashcard apps. Learners watch immersive clips, repeat phrases aloud, and review saved vocabulary through spaced repetition, an approach rooted in comprehensible-input research. The format mirrors social video feeds, turning passive scrolling time into structured language exposure.

Parse builds a web scraping and automation API that lets developers turn any website into structured, typed endpoints. Instead of driving a headless browser, Parse reverse-engineers a site's underlying network requests and generates API endpoints automatically, making it far faster and cheaper than traditional browser automation. Developers use it to extract data, submit forms, log in, and trigger actions on any website through a single API call.

Patent Watch builds AI software that automates patent infringement detection for companies and legal teams. It reads patents, surfaces competing products, and produces claim charts that map exactly which claims a product infringes and how, replacing the manual attorney work that typically costs tens of thousands of dollars per engagement. Teams use it to pursue licensing targets, send cease-and-desist letters, counter litigation, support M&A diligence, and run prior-art and invalidity searches.

Piggy Robotics builds humanoid robots designed for household chores, targeting everyday consumers rather than industrial operators. Instead of conventional electric motors, the robots use artificial muscles: flexible tubes wrapped in braided fibre and driven by a single pump, a construction method cheap enough to support mass production at scale. The goal is to put a physical robot in the home that handles routine domestic tasks a person would otherwise do themselves.

Pixley AI lets children and parents co-create personalized animated cartoons and storybooks by turning kids' drawings, photos, or verbal descriptions into voiced, animated episodes in about a minute. The product targets families frustrated with passive, low-quality streaming content, replacing professionally produced children's entertainment with stories where each child's own characters take the lead. Pixley's longer-term goal is a fully generated video platform where all content is built around each viewer rather than recommended from a fixed catalog.

Play Health builds a perimenopause care platform for women and their clinical providers, combining a patient app with a clinician dashboard. The platform aggregates symptom, medication, and lifestyle data to generate AI-driven insights and clinical protocols, replacing generic, one-size-fits-all guidance with personalized, evidence-based care. Each clinical interaction feeds a longitudinal dataset that informs better patient outcomes and positions the company to partner with pharma and biotech firms developing women's health therapeutics and diagnostics.

PlayVision builds AI-powered film analysis and player-ranking software for basketball coaches and scouts. Teams upload practice or game footage and the platform automatically tags plays, tracks frame-by-frame movement, and surfaces advanced analytics on speed, positioning, and decision-making, replacing hours of manual film breakdown. A separate transfer-portal database ranks college and professional players across more than 40 metrics, giving recruiting staff a structured alternative to traditional scouting reports.

Primer builds an AI agent that guides prospects and new users through live, personalized product walkthroughs, handling sales conversations, onboarding, and support without human involvement. It targets B2B software companies and replaces the work of sales engineers, onboarding specialists, and support reps who would otherwise run those sessions manually.

Prox builds AI agents for third-party logistics providers and fulfillment centers, the warehousing and shipping operations that brands outsource. Its digital workers handle back-office tasks, replacing the manual administrative labor that consumes large volumes of staff time across 3PL operations.

Questom builds AI agents that handle inbound B2B sales conversations for mid-market and enterprise companies. The agents answer calls and emails, ask qualification questions, pull pricing and catalog rules, generate quotes, and push structured orders directly into backend systems such as ERPs and order management tools. This replaces the manual, repetitive work that operations and sales teams do today across email, PDFs, spreadsheets, and business software.

Redapto builds continuous quality-improvement software for customer support teams. It audits every agent interaction to surface SOP gaps, hallucinations, and stale knowledge, then automatically updates procedures, training, and knowledge bases. This replaces manual QA programs that review only a small fraction of tickets and leave systemic issues undetected.

Remedy builds AI agents for pharmacies that handle prescription refill workflows, replacing the manual calls, follow-ups, and coordination tasks that pharmacy staff currently do by hand. The product targets independent and chain pharmacies looking to cut administrative load while keeping patients on their medications.

Rivet builds a visual AI editor that lets product designers make UI changes directly in production code, bypassing the handoff to engineers. Designers push changes straight to the codebase, eliminating the back-and-forth that causes their designs to diverge from what ships. The product targets design-led companies where the gap between design intent and coded output slows teams down.

Rovi Health builds an AI-powered clinical navigation platform for employer health plans. AI agents analyze member clinical data to surface care needs, treat common conditions through an integrated virtual clinic, and match members with high-quality in-network providers for complex care, handling appointment booking end to end. The product replaces fragmented benefit navigation, reactive care management, and manual provider search for both employers and their employees.

Rovr builds an AI front desk for engineering teams that handles incoming requests, routes them to the right owners, and posts status updates directly in Slack. It targets platform, dev-infra, and internal tooling teams, replacing manual Slack triage, internal intake forms, and scattered documentation. A built-in ops CRM gives the entire organization a single system of record for engineering requests, from access provisioning to incident coordination.

s2.dev builds a serverless API for durable, real-time streams, giving developers a cloud storage primitive for moving data the way object storage handles static data. Engineers use it to persist and read ordered, append-only streams at unlimited scale without managing clusters or brokers. It replaces self-hosted Kafka setups and ad-hoc log infrastructure for use cases like agent coordination, real-time data broadcasting, and multiplayer sync.

Sanctum builds an AI-powered user simulation platform for software teams shipping new features. It ingests real session recordings to construct behavioral user models, then runs those models against code changes to catch bugs, UX regressions, and performance problems before release. The product replaces manual QA scripting and end-to-end test authoring by generating and executing realistic user flows automatically.

Sava builds software infrastructure for trust creation and administration, serving trusts and estates attorneys, registered investment advisors, and family offices. It replaces manual, document-heavy workflows that professionals use to structure and manage wealth transfer through trusts.

Sciloop builds an AI co-scientist that automates the experimentation and analysis loop for machine learning researchers. It replaces the manual work of designing, running, and interpreting ML experiments, letting researchers test ideas at a pace that current tools cannot match. The target user is the ML researcher who today loses weeks to workflow friction that sits between a hypothesis and a result.

Scoop builds AI agents that automate the preparation and ongoing maintenance of IND submissions for biotech companies. It ingests scientific reports, CMC packages, and supporting documents, then generates submission-ready regulatory narratives with full traceability to source files and automatic detection of cross-module inconsistencies. The product replaces the months of manual document consolidation that regulatory affairs teams, CROs, and consultants currently perform to get a drug candidate to its first-in-human trial.

Scott AI builds a neutral decision workspace where engineering teams align on product specs before handing work off to coding agents. It replaces ad-hoc use of general-purpose docs tools like Google Docs and Notion as the primary place teams write and agree on specs. Because execution agents each favor their own tool, Scott AI sits above them as an agent-agnostic layer that coordinates what gets built and why.

Selfin builds an AI-powered consumer bank that consolidates a user's financial accounts, investments, and products into one platform. It then generates personalized recommendations to optimize how each user manages and allocates their money, replacing the manual work of monitoring multiple financial apps and making decisions across fragmented accounts.

SellRaze builds an AI-powered tool that lets individuals and resellers create and publish secondhand product listings across multiple marketplaces, including eBay, Poshmark, Depop, Mercari, and Facebook, in seconds. A seller points their camera at an item, and SellRaze identifies it, prices it, and generates the listing automatically, replacing the manual work of researching, writing, and cross-posting. The platform also centralizes inventory tracking, buyer communication, and shipping label creation across all connected channels.

Semble AI builds AI agents that autonomously design building systems, including fire suppression, HVAC, plumbing, and security, for commercial and residential construction projects. Its customers are building system integrators who today produce these designs manually through spreadsheets and hand-drawn schematics, a process that can stretch across weeks. Semble replaces that manual drafting and coordination work end to end.

Semiotic builds AI design agents that produce studio-grade landing pages for fast-growing startups. The product targets companies that need high-quality web design but cannot afford agency timelines or five-figure minimums, and lack the tools to produce polished work themselves. It replaces both expensive agency engagements and the low-quality output of existing AI website builders.

SF Tensor builds infrastructure that lets AI researchers train models without managing the underlying compute stack. Its platform spans two layers: a managed cloud that sources GPUs across providers and handles spot preemption, and an automatic kernel optimizer that tunes training code to the target hardware without requiring hand-written CUDA. The work it replaces is weeks spent by researchers configuring distributed systems, debugging infrastructure, and writing hardware-specific kernels instead of doing research.

Sorce builds a swipe-based job search platform where an AI agent reads a job seeker's resume once and then fills out applications, uploads documents, and answers screening questions across employer career sites on their behalf. It targets the full range of job seekers, from entry-level to executive, replacing the repetitive manual work of submitting applications one form at a time.

Sourcebot builds a self-hosted, open-source code understanding platform for large engineering teams and AI coding agents. Developers use it to search millions of lines of code with regex, navigate complex repositories, and get answers grounded in their full codebase, replacing the manual process of hunting through code or pulling senior engineers away to explain it. AI agents connect to Sourcebot to retrieve company-wide codebase context, reducing hallucinations caused by limited local file access.

Spatial AI builds data infrastructure for robot foundation models, collecting and structuring large-scale, real-world physical data that general-purpose robotic systems need to learn from. It serves robotics and AI developers who cannot source this training data on their own. The work it replaces is the manual, ad-hoc effort teams spend assembling embodied-agent datasets.

Specific builds a cloud platform that lets coding agents provision and manage backend infrastructure alongside the code they write. Developers connect their AI coding agent, which then defines resources, spins up services, and deploys to production without manual infrastructure configuration. Specific replaces the work of manually setting up and operating backends, covering services, databases, storage, and edge delivery.

Stratus Aviation builds a real-time intelligence layer for airport and airside operations, aggregating radio transmissions, aircraft movements, and sensor data into a single shared picture. It serves airport operators whose ground coordination today depends on manual logs, radio calls, and siloed systems that create blind spots across the airfield. Stratus replaces that fragmented, manual oversight with unified, live data that connects communication, movement, and intent.

Structured AI builds an AI-powered drawing review platform for architects and engineers on construction projects. Its agents check every sheet in a drawing set against project standards and building codes, catching coordination clashes, compliance failures, and inconsistencies that manual redline reviews routinely miss. The product replaces the end-of-deadline red-pen pass that has long been the industry's last line of defense before errors reach the field as costly RFIs, change orders, and rework.

Sylvian builds specialized AI agents for companies, delivering purpose-built agents trained for specific business tasks rather than general-purpose assistants. It targets B2B customers who need AI that can take on defined workflows end-to-end, replacing manual or rule-based processes with agents tuned to the job at hand.

Telemetron builds an AI-powered support and operations platform for hardware companies, covering medical devices, consumer electronics, industrial equipment, and robotics. It connects directly to devices in real time to diagnose issues, route tickets, and trigger resolutions automatically, replacing manual triage workflows and fragmented tooling across ticketing, diagnostics, order tracking, and documentation.

The Context Company builds observability tooling for AI agents running in production, serving engineering teams that need to find and fix failures their standard monitoring never surfaces. It tracks traces, tool calls, latency, and costs alongside semantic analysis that catches silent failures such as wrong tool selection, loops, and hallucinations. Developers instrument their agents in under ten lines of code and investigate issues directly from their IDE.

The Hog builds a web intelligence API that lets B2B sales and marketing teams find, enrich, and monitor prospects in real time. It replaces static lead databases and multi-vendor data stacks by continuously scanning live social conversations, news, and web signals to surface verified contact and company intelligence. A single operator can identify in-market buyers, enrich records, and trigger outreach across LinkedIn, Reddit, and X without manual research.

Thesis builds an AI-powered discovery engine that lets researchers and builders construct machine learning models to accelerate scientific and technical breakthroughs. The platform treats discovery as a systematic engineering problem rather than a trial-and-error process, replacing the slow, resource-intensive work of manual hypothesis generation and model development. The target users span science and industry teams seeking to reach results of the scale of AlphaFold or the Transformer architecture, faster and more reproducibly.

Tornyol builds micro-drones that autonomously detect and kill mosquitoes. The drones weigh roughly 40 grams and use smartphone microphones, parking-sensor hardware, and custom signal processing to locate and eliminate mosquitoes mid-air. The product targets urban mosquito control, replacing conventional abatement methods and addressing the spread of diseases such as malaria, dengue, and West Nile virus.

Unsiloed AI builds vision-model infrastructure that converts multimodal documents, containing text, tables, images, and charts, into structured, LLM-ready data via API. It targets AI engineering teams at startups and enterprises that would otherwise spend months hand-building document parsing and chunking pipelines, most of which never reach production. The API replaces that bespoke work and feeds clean, queryable output directly into RAG pipelines and downstream automation.

Uplane builds an AI system that replaces marketing agencies and their supporting stack of tools for B2B and consumer brands. It reads first-party CRM and ERP data alongside live market signals, generates up to a thousand brand-compliant ads and paired landing pages per month, deploys them across Meta, Google, TikTok, LinkedIn, and other channels, and continuously reallocates budget toward top performers. The work it replaces spans creative production, media buying, audience segmentation, and cross-channel reporting that companies previously split across agencies, spreadsheets, and disconnected platforms.

Velvet curates and delivers video datasets to frontier AI labs training world models and conversational AI systems. It sources high-value video from individuals across everyday and specialized domains, then verifies each dataset for quality, uniqueness, and compliance before shipping it to customers. The work it replaces is the ad-hoc, unstructured process labs would otherwise use to find, vet, and ingest raw video data at scale.

Veria Labs builds an autonomous penetration testing agent that continuously scans a company's attack surface, produces verified working exploits, and opens pull requests with ready-to-merge patches. It serves software companies that rely on periodic manual security audits or static scanners, replacing those point-in-time reviews with ongoing, automated testing that surfaces exploitable vulnerabilities in hours. The product integrates findings directly into Slack and Linear, keeping remediation status in sync without manual handoffs.

Wardstone builds a modular, mass-producible interceptor platform designed to defend against mid-tier threats, from Group 4 drones to medium-range ballistic missiles. The platform is sized and priced to match the threat, replacing the current reliance on a small number of costly, oversized interceptors that cannot keep pace with adversary production volumes. Each interceptor is assembled from interchangeable modules covering aerodynamic surfaces, propulsion, avionics, divert-and-attitude control, payload, and seeker.

Zalos builds AI computer agents that log into existing finance systems, such as ERPs and accounting platforms, and execute repetitive back-office work including reconciliation, billing, transaction categorization, and month-end close. Finance teams at companies using tools like NetSuite get rule-based processes automated without API integrations or system replacements. The agents handle authentication barriers, extract and harmonize data, and self-improve over time, replacing manual grunt work done by in-house or outsourced finance staff.

Zarna builds AI associates for private equity firms and investment banks working middle-market M&A. Its agents connect to a firm's existing data and systems to screen deal teasers, redline NDAs, populate CRMs, build LBO models, and draft IC memos, replacing the manual, low-judgment work typically done by junior deal team members. The result is that dealmakers spend less time on data entry and document processing and more time on relationships and decisions.

Zavo builds an AI-native operating system for restaurants and hospitality businesses, combining point of sale, payments, reservations, and AI automation into a single platform. It targets enterprise and multi-location operators, replacing a patchwork of separate tools for transactions, table management, and manual back-office work. The platform is customized to each operator's workflows rather than requiring the business to adapt to a generic system.

Zephyr Fusion builds a compact, megawatt-class magnetic confinement fusion reactor designed to operate in orbit, delivering power to space-based industries at a fraction of the cost of equivalent solar arrays. The reactor targets customers across microgravity manufacturing, in-space data centers, defense, transportation and propulsion, and human habitation, replacing the expensive, low-output solar infrastructure that currently constrains the space economy. The founders are fusion PhDs with hands-on reactor research experience at Oak Ridge National Laboratory and Lawrence Livermore National Laboratory.