ML Startups funded by Y Combinator (YC) 2026

Understudy is an inference cloud. We capture your AI work traces, train smaller models, and automatically deploy them (only when they outperform the expensive model you are currently paying for). Instead of renting the same frontier model forever, your AI infrastructure gets faster, smarter, and cheaper every time it runs. Suppose your product uses Claude for a repetitive legal workflow. Understudy observes successful production runs, learns the specific tools and judgment required for that workflow, and trains a specialized open-weight model. Once that model matches or beats Claude on a held-out evaluation, Understudy starts routing production traffic to it. As more work is completed, the model keeps improving. We started Understudy because we believe every company using AI should be accumulating intelligence, not accumulating API bills.

Expanse unlocks wasted GPU capacity. We recover idle compute through three capabilities: resource prediction (right-sizing job submissions before they reach the scheduler), optimisation suggestions (code and config changes researchers can apply themselves), and failure prediction (catching jobs that will fail before they consume hours of GPU time). We’re four engineers. We ran HPC and GPU training workloads at the largest quant funds and national supercomputing centres. We faced this problem first hand and the only fix was to over-provision and burn millions. Ismaeel built the first multimodal HPC resource predictor as research at EPCC (Edinburgh’s Parallel Computing Centre), which beat every published baseline. This is the tool we wish we had.

Ndea is building frontier AI systems that blend intuitive pattern recognition and formal reasoning into a unified architecture. # AI for Scientific Advancement Unlike all life before us, humanity's ascent is a story of ingenuity, not just biological evolution. Our progress has been driven by the curiosity to acquire knowledge, the ability to pass it on, and an intrinsic drive to innovate. We build technology which gives us leverage beyond our biology. We stand at the top of a knowledge and technology colossus that we collectively created over the past ten thousand generations. Scientific progress has helped us overcome burdens that long defined human life — famines, plagues, and widespread illiteracy. Science will continue to redefine the boundaries of the human condition. Today, the acceleration of scientific progress hinges on one factor: AI capable of independent invention and discovery. This capacity is the gateway to advancements beyond our wildest imagination. # A New Research Lab We're starting Ndea — an AI research and science lab. The name - like 'idea' with an 'n' - is inspired by the Greek concepts ennoia (intuitive understanding) and dianoia (logical reasoning), capturing our first goal to merge deep learning with program synthesis. Ndea is entirely focused on developing and operationalizing AGI to realize unprecedented scientific progress in our lifetime for the benefit of all current and future generations. Building AGI alone is a monumental undertaking, but our mission is even bigger. We're creating a factory for rapid scientific advancement — a factory capable of inventing and commercializing N ideas. From our vantage point today, we see many 'known' frontiers like self-driving vehicles, drug discovery, sustainable energy, robotics, and space exploration. While AGI will benefit all of these, the most exciting adventure lies in the 'unknown'. AGI promises discoveries and progress we cannot imagine today.

Datoric provides custom training data for voice models, robotics, and world models. We manage the full process from contributor recruitment and task design to secure collection, quality control, rights management, and delivery. Our collection infrastructure is private by design. Voice, robotics, and world-model projects run through separate applications and sites, so contributor pools, task requirements, incentives, and data flows do not mix across modalities. Within each system, customer projects are isolated in private workspaces. Only approved contributors can access assigned tasks, with no public task listings, upload links, or contributor APIs. We also build quality into the collection process instead of relying only on review after the data arrives. Contributors are qualified and matched before they participate, collection requirements are enforced inside each workflow, and early submissions are tested before a project scales. Every sample is then linked to its contributor, consent, usage rights, collection history, and review record. Datoric’s mission is to make human-generated training data more secure, reliable, and responsibly sourced from the moment it is created.

What if every experiment you wanted to run could be modeled using AI? Godela is an AI-powered physics engine that gives engineers faster, cheaper replacement to simulations and physical prototypes. No set up, no wait. Just ask—and get simulation quality result.

We're creating reinforcement learning environments for training frontier models.

Feyn is an end-to-end service for building custom models from your data. Our models learn how your product actually works, handle the rare cases that are costly to get wrong, and grow sharper the longer they run. Our engineers start by reviewing your current setup and agreeing on what quality means for your workflow. Then we train a model on your data. Once it ships, new cases feed back into training so the model keeps improving. The result is a model that runs faster, improves with use, and is yours to own.

RunLocal is an environment that specializes AI agents to optimize model for onboard compute platforms like NVIDIA Orin/Thor and Qualcomm → www.runlocal.ai RunLocal is built for Physical AI engineering teams across autonomous vehicles, robotics, smart cameras and more. It tracks every experiment and continuously refines experimentation data into an understanding of what drives performance on your target hardware. This environment means that a generic coding agent (e.g. Codex or Claude Code) can experiment and iterate better, faster and cheaper. With RunLocal, you hit performance targets faster and ship more optimized models – without hiring inference optimization specialists. We’re working with leaders in autonomous vehicles and robotics. We're backed by investors like Y Combinator and 468 Capital.

Pipeshift helps engineering teams run real-time inference in production. We offer optimized runtimes to meet latency/throughput SLAs, paired with infrastructure orchestration that auto-scales and routes workloads across clusters and regions at cost-effective rates.

Dragoneye helps devs build their custom video detection models in less than 5 minutes. These models can then be used to power apps and features that use images and videos. No more arduous process of annotating any training data or doing any machine learning work themselves. Try out our demo today at https://playground.dragoneye.ai/ !

Over 80% of maternal mortalities were preventable with earlier intervention. Currently, doctors miss >50% of moms who have life-threatening complications and wait until labor to start doing manual risk assessments, at which point it's too late. Elythea's proprietary ML models catch life-threatening obstetric complications (like postpartum hemorrhage, preeclampsia/eclampsia, emergency c-section, preterm labor, and more) as soon as the first visit. We automatically analyze relevant demographic and clinical risk factors, flag high-risk patients, and provide evidence-based interventions, giving doctors months in advance to intervene and prevent complications from ever happening. Our global clinical studies have shown that we can catch 3x more complications than clinical judgment 10x sooner.

Cair Health empowers billing companies, care management platforms, EHRs, offshore BPOs, provider groups, and specialty pharmacies with AI-driven RCM solutions. Our suite of AI Agents helps RCM teams streamline operations, reduce manual costs, and improve consistency—while maintaining full control over core workflows. Cair is trusted by 100+ practices across the country — including iconic VC-backed digital health startups, leading specialty EHRs, and some of the largest billing companies and BPOs Backed by Y Combinator, Foundation Capital, Soma Capital, Translink Capital & many others..

truemetrics helps logistics companies to find their customer's faster. By detecting and mapping entrances, we help GLS, a key logistics company in Europe, save 32.7 seconds on each delivery. The way couriers navigate is broken. “You Have Arrived At Your Destination!” — or so Google Maps thinks. In reality, the way couriers navigate is broken: No parking, hard-to-find entrances, and labyrinth-like apartment and office complexes can turn the last meters of delivery into a time-consuming puzzle. truemetrics provides parking positions and entrance locations to enable faster drop-offs. Our data is generated by analyzing sensor data from drivers' smartphones, allowing truemetrics to identify precise parking and entrance locations, thereby simplifying the last meters of delivery.

We're building an AI consumer underwriting automation agent.

Rownd AI builds fully autonomous drones that use local edge intelligence and global LTE/5G/SatCom failover to eliminate the 1:1 pilot bottleneck anywhere on earth or space. We are currently field-deployed with active customers across U.S. Special Operations Command (SOCOM) via our defense business line, ArcEdge AI. Operating out of Raleigh, NC and Wisconsin, we leveraged our team's background in mobile hardware optimization and zero-trust architectures to build true cognitive autonomy directly onto physical aircraft. While we are heavily focused on defense technology today, our ultimate mission is scaling this physical autonomous infrastructure across commercial enterprise edge applications.

Emobi is a unified legal, financial, and technical infrastructure that delivers seamless, secure, and reliable EV charging across networks—powering apps, fleets, and automakers while maximizing charger utilization. Our secure, fully automated charging technology, JustPlug, works instantly with no special hardware, software, or setup, making EV charging effortless. With 122,000+ EV chargers integrated across the U.S. and Canada, Emobi is recognized as the largest EV charging ecosystem in North America. Backed by two provisional patents and 15+ enterprise customers, Emobi's technology is also trusted by the Department of Energy (DoE) and the Department of Transportation (DoT). To learn more, visit www.emobi.ai.

Telmai is a Data observability and Data Quality platform designed for Open architecture. It monitors data stored/managed in any data system, such as a data lake or data warehouse, in its raw format at scale without sampling

Our platform fixes construction's broken supply chain. 70% of construction projects finish late or over budget; procuring and managing building materials is the number 1 reason why. We use LLMs to buy and manage building materials for large construction companies. With DigiBuild, construction contractors can find, order, track and manage their materials from the supplier to the job site. We help customers save money, staff time, and improve construction project schedules. We work with the top real estate developers, general contractors, and subcontractors in the US. We manage billions in building material volume annually in the 2nd largest market in the world.

Superb AI is a leading computer vision platform and professional services provider that provides enterprise-grade, end-to-end MLOps and DataOps workflows to accelerate the adoption and development of data-centric AI. Through the practical application of AI-based automation, Superb AI helps teams manage the entire ML lifecycle more efficiently, from data annotation to curation, model training, and deployment, while ensuring optimal data accuracy and consistency.

The artist-formerly-known-as Scoopler is a real-time search engine. It aggregates and organizes content being shared on the Internet as it happens, like eye-witness reports of breaking news, photos and videos from big events, and links to the hottest memes of the day. It does this by constantly indexing live updates from services including Twitter, Flickr, Digg, Delicious and more. When users search for a topic on Scoopler, they get the most relevant results, updated in real-time. Scoopler has now become JustSpotted, a celebrity-tracking tool, meant to give fans an idea of what their favorite celebrities are up to at any given moment. JustSpotted, like Scoopler, is a data-aggregator, yet JustSpotted collects from Twitter, Facebook, and Foursquare -- alongside their own data they’ll be collecting from users who actual spot celebrities in the wild. All the information we aggregate is publicly available and you can even do a search on Twitter yourself and find sightings of celebrities. We're just the first people to organize it on a map and in such a way that it's easy to consume and ENJOY.” JustSpotted is about connecting with celebs. All the sightings on our website are naturally occurring events where an everyday person bumps into a celebrity as a chance occurrence. Back in the day, when you saw a celebrity while you were out and about the first thing you would have done was phone or text a friend. In the new age of social media, the first thing you do is share it on Facebook or Twitter. This phenomena is already happening. The service, which launches the week of October 11, 2010, is the latest creation by AJ Asver, Dilan Jarawardane and Benjamin Tauber.