Y Combinator’s March 2026 machine learning startup directory lists 197 companies in the category, alongside a broader startup directory with over 5,000 companies. The group spans mature late-stage businesses and very early teams, with examples ranging from large platform companies such as Flock Safety and Scale AI to newly funded W2026 startups with 2 employees. The selection reflects machine learning’s reach across public safety, enterprise software, data infrastructure, robotics, biotech, healthcare, developer tooling, logistics, retail, and financial services.
Several companies focus on core infrastructure for Artificial Intelligence development and deployment. Scale AI describes a data-centric platform that uses RLHF to help organizations build stronger models. The Token Company is building compression middleware for natural language workloads, and says compressed requests lifted purchase volume by 5%. Velum Labs targets automated data quality enforcement, while Hyperspell builds memory systems for agents connected to workplace tools and notes experience scaling API products to $30M ARR. Lilac, Osmosis, Ocular AI, Preloop, Kashikoi, and mlop.ai also address the stack around inference, fine-tuning, annotation, deployment, benchmarking, and model operations.
Another large cluster applies machine learning to physical industries and robotics. One Robot builds simulation environments for robotics teams working on manipulation tasks such as textiles and box folding. DeepAware AI says its automation system for GPU-intensive data centers can slice energy waste by up to 30%. AutoPallet Robotics targets warehouse case picking and mixed-SKU palletization in a market where retailers in the US spend approximately $10B per year on human labor for moving boxes. K-Scale Labs, Photonium, Conductor Quantum, Moonshine, and Allus AI extend that focus into humanoid robotics, optical design, quantum automation, mapping, and manufacturing vision systems.
Healthcare, science, and research also stand out. Exonic is building biological foundation models for safer gene therapies and says that in 2025 it set a new state of the art in liver cancer targeted gene therapy. Sciloop is developing an Artificial Intelligence co-scientist for machine learning researchers. Mecha Health is building x-ray foundation models and says radiologists can move from reading 1 scan per hour to 1 scan every 5 minutes, with x-ray report generation representing a 40B+ market opportunity. Undermind targets expert search and says its approach outperforms traditional keyword search by 10-50x for complex research queries.
The directory also shows strong demand for machine learning in business operations and vertical software. Booko reports early customers see ~20% revenue uplift through dynamic pricing of bookable time slots. Anglera says it reduces time per product from 15 mins down to 5 seconds for e-commerce product data workflows. FINNY AI focuses on advisor growth and notes that advisors often waste 60 hours of business development to convert 1 new client. Artificial Societies says it has built over 2.5 million Artificial Intelligence personas, delivered over 18 million responses to global Fortune 100 enterprises, and informed over 100 million dollars’ worth of decisions. Together, the companies suggest Y Combinator’s machine learning cohort is less a single category than a broad operating layer for modern software and industry.
