Machine Learning Startups funded by Y Combinator (YC) 2026

Flock Safety provides the first public safety operating system that empowers private communities and law enforcement to work together to eliminate crime. We are committed to protecting human privacy and mitigating bias in policing with the development of best-in-class technology rooted in ethical design, which unites civilians and public servants in pursuit of a safer, more equitable society. Our Safety-as-a-Service approach includes affordable devices powered by LTE and solar that can be installed anywhere. Our technology detects and captures objective details, decodes evidence in real-time and delivers investigative leads into the hands of those who matter. While safety is a serious business, we are a supportive team that is optimizing the remote experience to create strong and fun relationships even when we are physically apart. Our flock of hard-working employees thrive in a positive and inclusive environment, where a bias towards action is rewarded. Flock Safety is headquartered in Atlanta and operates nationwide. We have raised $150M in our Series E led by Tiger Global at a $3.5B valuation.

Scale accelerates the development of AI within organizations of any size to deliver critical business insights and operational efficiency. Its data-centric infrastructure platform leverages RLHF (Reinforced Learning with Human Feedback) to help organizations build the strongest AI models that supercharge their business, with customers across industries including Meta, Microsoft, U.S. Army, DoD’s Defense Innovation Unit, Open AI, General Motors, Toyota Research Institute, Brex, Instacart and Flexport.

Compression middleware that removes context bloat in milliseconds, lowering costs and improving end-to-end latency. Compression is especially effective across natural language workloads. In a blind LLM arena case study with one of our customers, compressed requests increased user preference, lowered costs, and lifted purchase volume by 5%.

One Robot builds simulation environments that are realistic to see and realistic to interact with, so robotics teams can train and evaluate robot policies without being bottlenecked by robot time. Today, improving a VLA often means more real-world hours: setting up the scene, running trials, resetting, and repeating. This loop is slow, expensive, and hard to scale. For example, material handling and manufacturing assembly tasks, models need far more training and evaluation data than teams can collect in the real world. We use task-specific data to build world model-based simulation environments for hard manipulation tasks (for example, textiles and box folding). These environments help teams run more training and evals, find failure modes faster, and accelerate iteration on action policies with less dependence on real-world data collection and robot availability.

Librar Labs is the data and intelligence layer of literature. We're working to solve the biggest bottleneck for the AI singularity: the world is full of massive amounts of unstructured knowledge that is hard for computers and AI to navigate. Our first commercial product is for the literature industry with a focus on school libraries. It captures and analyze unstructured (and pre-Librar unattainable) data using it to more than double reading rates at schools. Fastest growing library tech company ever backed by founders and operators of OpenAI, Depict, google maps, kahoot etc.

Booko helps businesses that sell bookable time slots make more money by dynamically pricing their time. Unlike physical inventory, unsold time slots disappear forever, yet most time-slot-based businesses still rely on static pricing. We integrate with existing booking systems to adjust prices and incentives based on historical utilization, availability, and demand. Early customers see ~20% revenue uplift by selling time that would have otherwise gone unsold.

RamAIn can intelligently automate any super complex and repetitive process across applications (Browser & Desktop) super fast. It doesn’t matter whether you have API access or not - if you can see it, RamAIn can do it. Current CUAs all follow the same decision loop: screenshot, VLM, decision, repeat. This is expensive, unreliable and slow. We build infrastructure to pre-train CUAs on specific interfaces, effectively teaching RamAIn to use your computer.

We provide real-time video intelligence that identifies escalating incidents as they unfold. Protent detects early escalation patterns and predictive threat signals, giving surveillance staff the immediate context they need to prioritize alerts and intervene before situations spiral out of control.

Arzule is an AI-native ecosystem revenue platform for B2B SaaS teams, starting with partnerships. Arzule uses a quantitative approach to estimate ecosystem revenue impact by combining CRM deal outcomes with ecosystem signals such as shared customers, integrations, partner referrals, and co-selling activity. Using probabilistic regression models and influence chains across the ecosystem graph, Arzule estimates how much each partner or integration increases the likelihood of pipeline creation and deal conversion. This allows teams to prioritize the partnerships most likely to generate measurable revenue. Beyond attribution, Arzule structures and manages the entire partnership lifecycle from identifying the right partners, to prioritizing based on revenue potential, to activating and scaling those relationships. Using always-on AI agents and proprietary data, Arzule builds dynamic profiles of companies and predicts which opportunities will drive measurable revenue, when to engage, and how to structure each relationship for maximum impact. It also optimizes execution through automated outreach, partner coordination, and performance tracking, ensuring every partnership and ecosystem motion is tied to clear ROI.

Velum is the operating system for data quality. Velum automatically monitors and enforces data quality across a company's data stack, so bad data never reaches dashboards. We turn data quality from a manual task into infrastructure that runs itself. Data trust you can prove. From the pipeline to the boardroom.

What if you could train a biological foundation model on... the entire internet? Exonic is pioneering a new generation of biological foundation models, focused on heterogeneous, unstructured, and noisy datasets. Our first application is the design of safer gene therapies. In 2025, we used AI to set a new state of the art in liver cancer targeted gene therapy, validated in vitro in our lab in San Francisco. So far in 2026, we have trained a new model with unprecedented zero-shot generalization on genomic regulatory expression in hold-out biosample datasets. More to be shared soon.

Sciloop is building an AI Co-Scientist that automates experimentation and analysis for machine learning researchers, allowing them to test ideas far faster than today’s tools allow. We are Bilal and Osman, International Physics Olympiad medalists and the only Pakistanis admitted to MIT in our respective years, with several years of hands-on ML research experience inside MIT CSAIL. We have built large-scale training pipelines, worked with terabyte-scale datasets, and seen the research bottleneck firsthand while contributing to scientific ML projects. We left MIT to solve this problem directly and are now building a system that can eventually accelerate scientific discovery itself. https://sciloop.dev

Allus builds next-gen vision foundation models that bring real intelligence to manufacturing. Enabling factories to see, understand, and improve production in real time.

Hyperspell (YC F25) gives AI agents memory. It connects to tools like Slack, Gmail, Notion, and Drive so agents can recall, reason, and remember across company knowledge. Developers use Hyperspell’s API to turn disconnected agents into AI coworkers that understand their work. Conor and Manu founded Hyperspell after building their own workplace agent and seeing how limited AI is without memory. They have over 15 years of experience in machine learning and have scaled API-based products to $30M ARR.

Fixpoint is Uber for human data annotation. We’re building the network of experts and the platform to automate sourcing, vetting, and deploying them on AI training projects. It used to be a manual process to staff people on AI training data or eval projects - but Fixpoint leverages AI to staff people and create data better, cheaper, and faster. This includes a network of pre-vetted experts, AI-automated vetting, and annotator management + payroll.

Brickanta is building the AI-native operating system for construction, starting with the pre-construction workflows that determine the outcome of entire projects: bid analysis, cost estimation, and procurement. Our platform combines state-of-the-art AI with real project data, standards, and construction best practices. This enables construction, contracting, and real estate companies to identify gaps, risks and opportunities earlier, and to produce project and procurement documentation significantly faster and more efficiently.

Bezel helps fashion brands create virtual photo/video shoots with AI. Upload images of clothes, select the human you want to model it, and Bezel generates pictures and videos that rival a full studio production. Every detail of the clothing is rendered flawlessly. Try it for yourself.

OnDeck is the infrastructure layer that makes Vision Language Models accessible and scalable for enterprise. We let organizations instantly find any object, behavior or event, in any footage, without needing to train a model or collect any training data. The Pain: Creating vision models usually takes months: collecting training data, training, then deployment. Worse yet: + it’s often impossible to get enough data for a specific task, and + even the best cv models struggle to generalize across diverse camera setups, workflows and environments. To overcome these blockers, we bet early on the power of VLMs and built a vision engine that can generalize across any task and doesn’t need any training data. We published a NeurIPS workshop paper showing our new methods with VLMs beat traditional CV even at niche tasks. Our current customers include: - National Defense Organizations - Robotics Research - Security cameras - Behaviour analysis for port monitoring - Off-shore oil & gas monitoring

We are a full-service sell-side residential brokerage that lists and markets your properties. We also screen and schedule tenants showings with our AI agent for a third of the normal commission.

Lilac taps idle GPUs from cloud providers and enterprises, giving startups and researchers cheaper inference while letting companies monetize unused capacity. Fully automated with Kubernetes integration.

DeepAware builds an AI-driven automation system for GPU-intensive data centers. Our reinforcement-learning scheduler, real-time market integration, and unified dashboard slice energy waste by up to 30%. Coming soon: autonomous “robot-hand” inspections and maintenance to enable 24/7 operations with minimal staff. More info: https://deepawareai.com/ Jobs/Internship: https://www.deepawareai.com/careers

Kashikoi is a simulation engine to benchmark AI agents. We generate CPU friendly world models that autonomously interview agents and generate deep behavioral assessments. We built a similar technology at Moveworks which was used to ship 250+ enterprise agents to customers daily.

Photonium is building software to automate optical system design. We supercharge optics experts with intelligent tooling to reduce costs and deliver faster. We handle the full design stack — from optimization, verification, sourcing, to prototyping — for AR/VR, quantum, biotech, metrology/chip fab, LiDAR, and more.

Theorem is training models that make program verification 10,000 times faster. Using verification as a feedback loop, developers have found zero-days in GPU accelerated code and cryptography implementations, and sped up code migration in legacy systems. If you have complicated code that needs to be correct and secure, sign up for our beta!

A fully open-source, performant and actionable Weights & Biases alternative that saves you money. Check out our story here! https://www.reddit.com/r/mlopai/comments/1kkc9jp/mlopai_an_efficient_free_and_opensource/

Plexe builds predictive ML models from a problem description. It connects to data sources, conducts experiments, evaluates and deploys the models to an API endpoint.

AI models are terrible in non-English languages because it's nearly impossible to find training data in other languages. So, we're building the world's largest and highest-quality multilingual data library.

Artificial Societies simulates how large groups of people respond to opinion surveys and react to information. We help global F100 enterprises anticipate how their most important stakeholders react to their most important decisions: from how investors and opinion-leaders react to comms messaging, to how high-value customers react to marketing strategies. We are experts in human behaviour. James, our Founder and CEO, is a Cambridge psychologist and data scientist who led the seminal paper that studied how 33,000 AI chatbots interact. Patrick, our Founder and Chief Product Officer, is an Applied Behavioural Scientist with years of experience helping F500 enterprises conduct market research. They are joined by a team of behavioural scientists, social scientists, and political scientists, all with a passion for uniting technology and an understanding of humanity. We know that every public-facing enterprise decision has million-dollar consequences. From how to position towards opinion-leaders, industry peers, and policymakers, to which marketing content to invest in, knowing how stakeholders would react matters. But traditional market research methods are too slow, too expensive, and often just fail to reach the stakeholders that truly matter. We spend our lives solving this problem. Our offering is that we build bespoke AI simulations of high-value audiences, for insights on questions no one else can answer. We’ve built up over 2.5 million AI personas that are all grounded in real human behaviour – not just what people say, but also what they do. But our enterprise partners choose us not just because we have achieved 95% accuracy in simulating human opinions compared to human self-replication - more importantly, they choose us because of our ability to deliver impossible research projects. Such as having insights in under 24 hours, such as being able to test sensitive strategies on high-value audiences with 100% security, without human risk. As a result, we’ve delivered over 18 million responses to global Fortune 100 enterprises, that helped shape over 100 million dollars’ worth of decisions ranging from global expansion strategies, product positioning, advertising, and strategic communications. Our vision of Artificial Societies is a Societal World Model that enables infinite experimentation, so that every organisation can have awareness of outcomes, before making a decision. No planes fly without a wind-tunnel; no medicines are approved without clinical trials; and yet, the biggest decision societies face today are taken as bets. We see Artificial Societies as a historically inevitable technology, and see it as our lives' mission to bring it to life.

Osmosis helps companies use reinforcement learning to fine-tune open source models that outperform foundation models.

Mecha Health builds foundation models to automate x-ray analysis for radiologists. We take medical images and process them using proprietary models to produce accurate draft medical reports. Our first model was built in less than two months, and beat Microsoft, Google, and OpenAI on clinical accuracy metrics. On top of that, it’s two orders of magnitude smaller and trained with a quarter of the data. We are partnering with the largest privately owned radiology practice in the US and a multinational tele-radiology company to provide them with their own foundation model, enabling their radiologists to go from reading 1 scan per hour to 1 scan every 5 minutes. By charging on a per scan basis, x-ray report generation represents a 40B+ market opportunity.

We are building state-of-the-art SLAM, perception, and spatial understanding models that leverage vision alone to map, interpret, and interact with the world.

Metreecs helps retailers plan, buy, and allocate products using AI-demand forecasting. We prevent overstock and out-of-stock situations, allowing clients to eliminate waste, free up capital, and drive higher sales.

We're building Matcha, the faster, smarter way to hire clinical talent—helping hospitals find active, qualified candidates at a fraction of the time and cost of job boards or headhunters. Matcha engages candidates at scale, matching them to your organization by qualifications and cultural fit. Recruiters save time, hospitals save money, and candidates get a better experience—a radically better hiring model for everyone involved.

With AI increasingly automating away code generation, software engineers will spend more time reading, judging, and architecting code rather than writing it. Storia is building an open-source copilot that knows a company's codebase and its context. We are starting with Sage, a Perplexity-like agent for helping developers understand, judge, and generate software. Given an existing codebase, developers can ask Sage questions such as: 1) Given my project’s SLA and latency constraints, what is the appropriate underlying vector database to use? How would I incorporate it into my existing codebase? 2) Why should I pick Redis over Milvus as my underlying vector store? 3) Does this codebase in our organization still work and what steps are required for a complex integration with another library? Sage’s answers are directly supported by documentation and external references like GitHub, Stack Overflow, technical design documents, and project management software, preventing hallucinations. Today, Sage has up-to-date knowledge about open-source repositories (indexed daily). Tomorrow it will have a deep understanding of every line of code on the Internet. For teams, Sage will know about your private codebase too. No group has yet solved how to build an AI system that comprehends a codebase and its context and can empower every developer to architect better code, faster. This requires new research advances because vanilla RAG and out-of-the-box LLMs aren’t going to cut it. We have 20+ years of software engineering and AI research experience. Julia worked on precursors of Gemini using contextual neural techniques before they were called “RAG” (and applied it to products like Google Keyboard and Pixel phones). Mihail built the earliest LLMs at Amazon Alexa and launched the first contextual deep learning conversational AI system in production at Alexa.

Simplex builds web agents that companies use to integrate with legacy portals. Customers use Simplex to dispatch freight shipments, download customers’ invoices, fetch websites’ internal APIs, and more.

We’re building the next generation of warehouse robotics. In the US today, retailers spend approximately $10B per year paying human laborers to pick up and move cardboard boxes in warehouses. Existing solutions for automating this are expensive and difficult to install, which is why manual operation is still so prevalent. Our solution is different. We make swarms of small mobile robots that install into existing warehouses to provide a low-cost and robust automation solution for case picking and mixed-SKU palletization. Our novel technology allows these robots to be installed and operate at significantly lower cost than existing solutions while being both flexible and robust.

Cloudglue APIs (https://cloudglue.dev) makes it easy to transform video and audio into structured data for your LLM/RAG/AI application. Our developer-friendly APIs makes it easy to integrate video chat into your AI assistant with minimal heavy-lifting.

MinusX is a chrome extension that adds a side chat to your analytics apps (Jupyter, Metabase, Grafana, Tableau, etc). Given an instruction, our agent operates your apps - by clicking & typing, just like you do - to analyze data and answer queries. We believe an AI Data Scientist is a scientist, not yet-another-new-analytics-platform. MinusX interoperates with you in tools you already love and use, and as a matter of philosophy, gets out of the way.

At Anglera, we're developing a suite of AI agents to help e-commerce companies run their operations more efficiently. Our flagship agent helps our customers onboard, enrich, and manage their product data, reducing time per product from 15 mins down to 5 seconds. We previously developed ML to automatically enrich millions of products at Uber Eats, and we're now on a mission to automate the most common manual workflows for every e-commerce business.

At Undermind, we're building a search engine that can handle extremely complex questions. It’s geared at experts, like research scientists and doctors, who need to find very specific resources to solve high-stakes problems. We’ve rebuilt search from the ground up to address this. Our new approach employs high-quality LLMs to adaptively explore a database, mimicking how a human researcher carefully discovers information. This approach dramatically outperforms (by 10-50x) traditional keyword search and other modern AI-based retrieval methods. Our first target users are the 50 million researchers searching for scientific literature on PubMed and Google Scholar every month. We’ve have paying users across fields like medicine, ML, biotech, finance, and more.

Quantum computers will allow humanity to understand the world at its most fundamental level, enabling the acceleration of drug discovery and the development of new materials. Currently, quantum engineers spend days, sometimes weeks, manually getting their silicon quantum chips to operational conditions to realize two qubits. A qubit is the information-carrying unit of a quantum computer, analogous to a bit in a classical computer. We need billions of qubits to make a useful quantum computer. Therefore, automation software will be vital to realizing this goal. Conductor Quantum will develop AI software to remove the human from the loop, enable the scaling of quantum technology.

Cartage is the future of freight coordination. Transparent, tech-driven and eliminating the need for human coordinators.

FINNY is building a new operating system for independent financial advisors, and starting by helping them grow. Our goal is to empower the best financial advisors (the independents!) to serve all Americans who want financial advice. We are starting with growth, which is the most existential problem advisors face. The organic growth process is highly inefficient and full of noise. Advisors often waste 60 hours of business development to convert 1 new client. At FINNY, we are certain we can do better with AI. With our tool, advisors can: - Identify prospects within their target niche, aggregating thousands of data points per lead - Prioritize prospects based on their predicted likelihood of converting, a score unique to each advisor and prospect pair - Automate the outreach and meeting scheduling with their high priority prospects In essence, doing all the work for the advisors, and letting them focus on what actually matters, the client relationship.

We're building humanoid robots to do most of what you find boring or tedious. We have an open-source design which we are releasing to the public, which is capable of walking, talking and manipulating objects.

Ryse is the only marketplace where investors who want to buy leases can trade with real estate operators who want to sell leases.

Ellipsis will help your team merge code 13% faster. Ellipsis is an AI developer tool that automatically reviews code and fixes bugs on pull requests. It uses LLM agents to catch logical errors, security issues, and can even enforce a team's style guide. The coolest part is that after Ellipsis identifies an issue, developers can tag @ellipsis-dev to have Ellipsis implement the fix. Internally, Ellipsis actually executes the code it generates, just like a human does. As a result, we allow developers to generate working, tested code directly from GitHub/GitLab.

Ocular AI is the data annotation engine for Generative AI, Computer Vision, and Enterprise AI models. We help you transform unstructured, multi-modal data into golden datasets to power generative AI, frontier models, and computer vision. Ocular Foundry is the most intuitive, data-centric, and fastest platform that lets you label, annotate, version, and deploy your data for training models. It also orchestrates your annotation jobs, improving collaboration with members and annotators. With Ocular Bolt, shift from humans in the loop to experts in the loop to supercharge your data labeling and annotation projects. Our global expert workforce ensures fast, accurate results—no matter the scale or complexity of your data. Companies spend huge amounts on training data, but Foundry and Bolt are AI-native tools that lower costs, reduce manual effort, and accelerate high-quality data collection. We’re replacing outdated, clunky, and expensive data software!

Only 2 out of 10 ML models make it from experiment to production. Preloop helps automate the process of deployment, helping companies realize more value from their machine learning teams, while focusing teams' attention on science instead of engineering.