Deep Learning Startups funded by Y Combinator (YC) 2026

Shire is an AI floor manager for full-service restaurants, running on the cameras they already have. It sees what a great manager sees: a guest ready to order, a table waiting on the check, a four-top that needs clearing, a server that is double booked. Then it acts, alerting staff in the moment and seating each new party with the server who has room. A POS only knows what gets typed into it. Shire knows what is happening in the room, so wait-time quotes, staffing, and pricing run on real turn times.

The world generates petabytes of video every day, and almost none of it is readable to machines. Frontier VLMs that can unlock this (like Gemini) are too expensive to run at the volume real industries have, too generic for any single task, and can't be deployed on-prem where sensitive video lives. We are building an auto vision factory: take video plus a customer's objective, distill a frontier model, and post-train a small open model that is SOTA on that one task and serve it blazingly fast. Our first product, Marlin, is a 2B-parameter open-weights video VLM that matches Gemini-2.5-flash on dense captioning and temporal grounding at a fraction of the cost.

DeepMark is the authentication layer for AI voice. When an AI agent speaks, we embed a compact, machine-readable ID directly into the audio, not in metadata, which gets stripped the moment a file is touched. The ID resolves to a trust record, so a bank, a call center, or the person on the line can verify in real time who the agent is.

At Talos, we are building better, more modern predictive maintenance for the power industry with a focus on large transformers. Our devices continuously monitor critical power equipment and detect problems early on using anomaly-detection models. For each fault, we provide detailed analyses of the problem, exactly where and why it happened, as well as next steps teams should take.

Models are trained on the entire internet, but most of what happens on Earth is never written down. We turn those signals into training data so labs can make models better at science

We builds foundational AI models for chemistry. Almost nothing you buy is a single chemical. Products like datacenter coolant, GPU thermal paste, paint, and glue are formulations, which are mixtures of around ten ingredients each, and demand for them has grown quickly because of AI and EVs. Designing a new formulation is essentially a search problem, because there are more possible mixtures than anyone could ever test, and none of them can be predicted in advance. As a result, bringing a single formulation to market today takes three to five years of trial and error in a lab. Our models predict how a mixture will behave far faster than the standard methods used today, which lets us design new formulations in weeks rather than years, at a fraction of the cost. To prove this, we used them to invent three chemicals that never existed before, including a PFAS-free coolant for datacenters, and we have since synthesized and patented all three. We are already working with some of the largest manufacturers to design formulations for them, and we are building our own lab to generate the data that will train our next models.

We're scaling self-improving AI scientist teams to ideate/experiment/discover, generating new knowledge autonomously. Their discoveries reduced ARC-1/2 costs by 10,000× while maintaining state-of-the-art accuracy and yielded material advances on combinatorial ML problems. We believe that AI is bounded not by compute, but ideas. Scaling to millions of sessions of discovery enables leveraging its own discoveries and presents a novel training source capable of ushering a new era of self-reliant foundational models.

Edviro builds AI that operates energy infrastructure. We connect to meters, HVAC systems, batteries, solar, and other energy equipment and continuously handle the work required to keep them efficient: detecting and diagnosing problems to forecasting demand, simulating fixes, creating work orders, coordinating maintenance, and controlling equipment where authorized. Our energy world models learn how each site behaves, choose the highest-value actions, and verify the results against real operational, meter, and billing data.

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.

Ashr Manifold is a self-contained model training, hosting, observability, and continual learning platform for purpose-built open-weight models. Startups can use Manifold to post-train and manage fleets of continuously improving open-weight models without having a dedicated research team or infrastructure. Ultimately, this reduces operating costs and allows a greater degree of control over proprietary company knowledge. Enterprises can use Manifold to encode institutional judgements and priorities into open-weight models, reduce token costs, and manage every level of the AI life cycle, from data curation, to model serving on weights and infrastructure they own.

Robotic factories that build robotics in the US. Currently operating a 12,000 sqft factory in the Bay Area and actively manufacturing customer orders for some of the largest robotics companies in the US and the International Space Station. We’re a group of Berkeley graduate robotics researchers who previously won a full scale autonomous IndyCar competition at 160mph.

We are scaling foundation models that are curious about the world. We’ve seen general intelligence appear exactly once, and its hallmark has been curiosity: a dissatisfaction with what we already know about the world, an ability to learn from world experience, and a disposition to share knowledge with other individuals. Across generations, these human traits have built modern science, technology, and culture. We think superintelligence will come from models that work the same way: learning from their own observation, updating as they go, growing what they know. These models will start from what humanity knows and discover past us. We believe that scaling curious intelligence is a defining problem of our time. If solved, we will live to see a new era of technological advances, scientific understanding, and human flourishing.

Flywheel AI converts any existing excavators for contractors to enable remote ops to increase safety and productivity, and use robotics context dataset to train autonomous policies.

AutoComputer is a desktop robotic process automation system.  Given just a text prompt, our AI automates tedious tasks such as financial data entry by performing all the clicks and keystrokes for you.

We are building the next generation of deep-learning models for enzyme design to slash the cost of chemical manufacturing. The $6 trillion chemical industry is flawed: It produces 20% of industrial greenhouse gases, and is responsible for 15% of global energy usage. Enzymes offer a far more sustainable alternative to chemical synthesis and have already revolutionised how a select few chemicals are produced. The problem is each enzyme takes years of trial and error to develop. Our enzyme models learn the principles of catalysis, allowing us to design enzymes for each reaction, in days not years.

Anthrogen is engineering post-modality biology. Today's modalities reflect historical contingencies in biological progress, not fundamental categories. We develop the AI systems that design modular biological machines and the experimental infrastructure to instantiate them.

Automorphic has invented a way to infuse knowledge into LLMs via fine-tuning (surpassing context window limitations), enabling developers to rapidly iterate on and successively improve custom models cheaply and efficiently.

Cedana (YC S23) brings hyperscaler and frontier-lab orchestration capabilities for AI workflows. Our core capability is live migration for CPUs and GPUs workloads. This increases cost savings up to 80%, accelerates time to first token 2-10x, and enables stateful reliability of training jobs even through catastrophic GPU failures. We've integrated our solution into K8s, and support Kueue and Slurm for training distributed jobs, and Kserve for serving inference.   OpenAI, Meta and Microsoft have flavors of these capabilities internally and we’re bringing them to everyone.  Our vision is to transform cloud compute into a real-time, arbitraged commodity.  https://www.cedana.ai

Diffuse is building generative AI for protein design. Our mission is to build AI systems that engineer new and useful proteins with unprecedented control and accuracy. Our team has been behind breakthroughs in AI protein design for the past 7 years, including the first experimental validation of AI-generated proteins and diffusion models for protein structure and sequence.

Here at Spellbrush, we're passionate about making a good anime game. We also happen to be the world's leading generative AI studio — we're the team behind niji・journey. We are currently investigating how AI can be used to help human artists perform masterpieces in the most complex medium of our times: video games. Our games are characterized by a no-compromise approach to well-balanced gameplay married to a truthful love of visual arts.

Cerrion helps manufacturers automatically detect, understand and eliminate problems on their production lines using video-based Computer Vision. Our AI leverages standard CCTV cameras and learns how a manufacturing process looks like when things are going well and can automatically detect and track problems in real-time. For example, one of our customers, a Pepsi supplier producing 500 bottles per minute now automatically detects and reacts to a fallen bottle before it starts blocking their production line.

Documents are everywhere. Through all industries, understanding their content is a critical step of many processes that software builders throughout the globe want to automate. We focus on the science, the AI, the deep learning to give these builders the one API they need to automate the understanding of documents and give their software human-like superpowers. Based on three pillars: - A product: a universal self-service platform where builders can train their own models - A catalog of use-cases: We push critical use cases to the limits of human performances in critical markets such as Accounting software, AP Automation, Expense Management, KYC and many others - A internal capability to make tailor-made AI: We answer to specific needs from large customers that needs a custom AI of their own to handle their use cases

Biodock's cloud platform accelerates microscopy analysis, automating months of microscopy analysis and infrastructure to minutes with our end-to-end AI architecture. Scientists enjoy auto-scaling storage, GPU compute, and 30-50% more accurate analysis. We're building an amazing experience to translate microscopy images to therapeutic insights for academic and enterprise scientists.

Hey, thanks for reading! We founded Anima out of a very personal problem. Again and again, we saw people dying because they got bad care plans, weeks or months late. We build Care Enablement for care teams - combining online consultation with productivity tools in a realtime multiplayer dashboard. By doing this, we get people optimal care within 24 hours.

AI-native Robotics and Industrial Automation.

Handl converts documents (invoices, income stabs, custom forms, etc.) into structured data through simple API. 100% automation implies no need for any supervision/manual work on your side and allows you to reduce costs, improve response time, and operate accurate data. Powered by the merge of ML/AI + humans-in-the-loop to work in real-time even for the most complicated cases

We analyze thousands of video streams to find and track people without facial recognition. Our tech is available as an API and has multiple use cases. Unique people counting, forensic people search, falsa alarm filtering and many more.

Sapling offers an API and SDK to help businesses integrate language models into their applications. Its messaging assistant sits on top of CRMs and messaging platforms to help users more efficiently compose responses.

Here’s a secret between you and me: even the world’s largest manufacturers, companies like Tesla and Toyota, waste billions of dollars every year making products with quality issues. Building high-quality things at scale is incredibly hard. It doesn’t just happen because you hire smart people or buy good machines. It requires seeing problems early, understanding them deeply, and acting in real time, something factories were never designed to do. At Overview.ai, we’re changing that. We build custom hardware, edge AI, and software systems that give manufacturers real visibility into how their products are actually being made. Our technology helps catch defects earlier, reduce waste, and fundamentally improve how factories operate. This work matters, not just for our customers, but for keeping American manufacturing competitive in a world that’s moving faster every year.

We provide a simple API for creating, storing, versioning, and collaborating on multi-modal AI datasets of any size. With Activeloop's open-core stack, you can rapidly transform and stream data while training models at scale. Deep Lake powers foundational model training by acting as a vector database with significant benefits, such as (1) the ability to use multi-modal datasets to fine-tune your own LLM models, (2) storing both the embeddings and the original data with automatic version control, so no embedding re-computation is needed (3) truly serverless service with no vendor lock-in. How cool is that? GitHub loves us - we're one of the fastest-growing libraries there, and we're used by little-known companies like Google, Waymo, and Intel. No big deal. Our founding team hails from places like Princeton, Stanford, Google, and Tesla, and we're backed by Y Combinator & other Silicon Valley heavyweights. Activeloop is hiring, and we want you! Check out our open roles on our YC page and join the fun. 10-min demo: https://activeloop.wistia.com/medias/aibvo0dst2 Whitepaper: https://www.deeplake.ai/whitepaper

Macromoltek: Revolutionizing antibody design. Description: Macromoltek, a computational de novo drug design company, rapidly produces accurate and credible antibody designs. We have built a proprietary platform that enables design against difficult targets inaccessible by traditional methods and have are already designing antibodies for large biopharmas and smaller biotechs.

D-ID enables creators and developers to generate realistic high-quality AI personas easily and ethically through the use of our platform and APIs, based on deep-learning and AI-powered technology - Enabling Creative Reality™. D-ID is a Tel Aviv-based Creative Reality™ startup specializing in patented video reenactment technology using AI and deep learning. Established in 2017, D-ID created the first facial image de-identification solution to protect images and videos from facial recognition software. D-ID's products range from animating still photos to facilitating high-quality video productions and creating viral user experiences.

Welcome to Netomi AI, where we are revolutionizing the world of customer experiences through cutting-edge artificial intelligence. Our mission is clear: to create AI that not only solves mission-critical problems for the world's largest brands but also fosters genuine customer love. Backed by industry titans like Y-Combinator, Index Ventures, Jeffrey Katzenberg, and Greg Brockman, Netomi AI is at the forefront of defining the future of AI-driven customer engagement. At Netomi, we embody our core values in everything we do: Passion: We love what we do, genuinely. We persevere, creatively problem solve, and grow from adversity. Customer Focus: We obsess over creating a positive impact and deliver experiences that meet or exceed the needs of our customers, and their customers. One Team: We believe in working collaboratively as One Team to meet shared objectives and goals. Joining Netomi AI means being part of a dynamic, fast-growing team that values innovation, creativity, and hard work. As a key player in the Generative AI revolution, you'll have the opportunity to significantly impact our success while developing your skills and career in the ever-evolving field of AI. Netomi AI is not just a workplace; it's a community of visionaries shaping the future of customer engagement. If you're ready to be part of something extraordinary, where your passion aligns with our values and vision, we invite you to explore opportunities with us. Your journey to redefine customer experiences with brand-safe AI starts here. Mission To empower the highest quality customer experiences with brand-safe AI. Vision Building Brand and Customer Love

Focal Systems is on a mission to lower the cost of living for all mankind by automating and optimizing Brick and Mortar Retail with the latest advancements in AI. Focal Systems is the industry leader in retail automation solutions. By digitizing store shelves hourly and unleashing FocalOS, retailers unlock huge operational efficiencies, optimized merchandising, and streamlined supply chains which deliver impactful financial results. Focal is transforming retail by empowering store management to make automated, data-driven decisions. We are the operating system of retail.

Mashgin creates better retail experiences through visual automation. We’ve built a self-checkout kiosk that uses computer vision to scan multiple items without barcodes, reducing checkout time by 10x. We’re completely recreating the checkout experience in an industry that’s had little innovation in decades. Our clients see dramatic reductions in lines and revenue increases of as much as 400% as a result.

Numerion Labs is an AI-native company accelerating the discovery of life-saving medicines through the development and use of cutting-edge machine learning algorithms. The company unites computational chemistry, structural biology, and medicinal chemistry to pioneer the next generation of AI-driven drug discovery platforms.

For the first time in history, technology exists to create a healthcare system that anticipates your needs, responds with precision and is accessible to everyone -- regardless of financial means or geography. Since 2015, Akido Labs singular focus has been to make this vision a reality.

[Segments.ai](http://segments.ai/) is helping robotics and automotive companies label their multi-sensor data for AI training and validation. Our platform enables customers to efficiently annotate their point cloud and image data, accelerating their path to autonomy. We're a fast-growing, remote-first YC startup with a lean team and healthy runway. [Segments.ai](http://segments.ai/) is used by large organizations as well as innovative startups building the next generation of autonomous drones, delivery robots, self-driving cars, and more.

ML models are only as good as the datasets they're trained on, and that means that most improvement to model performance comes from improvement to the quality and diversity of their datasets. Our tooling makes it easy for ML teams to find anomalies + failure patterns in their datasets and fix these problems by editing / adding the right data. So the next time you retrain your model, it just gets better.

We help teams that may or may not have machine learning expertise quickly turn their data into deployed computer vision models. Example applications include security camera monitoring, automating data entry, interpreting web scraped images, validated user photo inputs, augmented reality and mobile product scanning. If you have visual or scanned data that you wish your software could interpret at scale, we can turn around a first proof of concept in under a week. Reach out and see how computer vision can change your business.

Formerly known as Patchd Medical, now acquired by Cytovale. Wearable technology and AI to predict and prevent sepsis outside of hospital.