Asimov
Product & Competitive Intelligence
Builds the training data supply chain for humanoid robots from real-world human movement.
Company Overview
Builds the data infrastructure for humanoid robotics. Workers wear a phone on a lightweight headband while performing daily tasks (cooking, cleaning, organizing), capturing egocentric video that provides the diverse, real-world human movement data that robots need to learn from.
Competitive Advantage & Moat
Product Roadmap & Public Announcements
Lightweight headband hardware, mobile recording app, and global partnerships with businesses. Hiring Data Collection Specialists worldwide at $5-30/hr. Already supplying data to major robotics players. Partners can cover worker salaries by having employees wear collection kits during normal work.
Signals & Private Analysis
Active global hiring for data collectors indicates rapid geographic scaling. Both founders are current UC Berkeley undergrads (graduating 2026). Lyem was founding engineer at Blume (YC W24) and did ML for Air Force fleet analytics at FLIP.
Product Roadmap Priorities
Crowdsourced Egocentric Video Collection & 3D Pose Annotation for Robot Imitation Learning
Workers wear lightweight headbands while doing their normal jobs, and the video gets turned into 3D movement recipes that teach robots how to fold towels, wash dishes, and navigate cluttered rooms.
It's like hiring thousands of cooking show hosts to wear GoPros while making dinner, then turning all that footage into a cookbook that robots can actually follow.
Motion Foundation Model Pre-training from Diverse Human Activity Data
Asimov is building a massive library of how humans move through everyday life so they can train one giant AI brain that lets any robot learn new tasks almost instantly.
Think of it as teaching a robot the "common sense" of physical movement the same way GPT learned common sense about language—by reading (or in this case, watching) everything humans do.
Automated Data Quality Scoring and Anomaly Detection Across Distributed Collection Network
An AI watchdog automatically checks every video workers upload, instantly flagging blurry footage, blocked cameras, or missing movements so nothing unusable makes it into the final dataset.
It's like having a strict but fair film editor who watches every take in real time and immediately yells "cut!" if someone left the lens cap on.
Company Overview
Key Team Members
- Anshul Verma, Co-Founder
- Lyem Ningthou, Co-Founder
Founder Force Multiplier
Anshul Verma brings data infrastructure experience from Scale AI internship, Amazon internship, and ML research at Berkeley BAIR. Lyem Ningthou was founding engineer at Blume (YC W24) and built ML data pipelines for the United States Air Force at FLIP. Former roommates who previously co-founded a startup to six figures in revenue.
Funding History
- 2026 | Anshul Verma and Lyem Ningthou co-found Asimov.
- 2026 | Accepted into Y Combinator W26 batch.
- 2026 | Launched headband hardware and global data collection operation.
Competitors
- Data Providers: Scale AI, Open-X Embodiment, Hugging Face LeRobot.
- Motion Capture: Xsens, Rokoko.
- Synthetic Data: NVIDIA Isaac GR00T Blueprint.
- Robotics Internal Data: Physical Intelligence, Figure AI, AgiBot.
