NomadicML (operating as Nomadic) is a San Francisco-based startup founded in 2022 by Mustafa Bal (CEO) and Varun Krishnan (CTO), both Harvard computer science graduates. The platform was originally positioned as an open-source compound AI system optimization tool focused on hyperparameter tuning for LLMs, RAGs, and guardrails, but has since pivoted to focus on physical AI infrastructure for robotics and autonomous vehicle teams. As of 2026, NomadicML functions as the spatial intelligence and data management layer for physical AI. It ingests video footage and sensor data from vehicle fleets and robotic systems, processes it through a collection of vision language models, and converts raw archives into structured, searchable datasets. This solves a core problem in the AV and robotics industry where up to 95% of fleet-captured data sits unused in archives. The platform supports natural language search over footage to find complex edge-case sequences in seconds, automates video labeling and event detection, and generates training-ready datasets for reinforcement learning. Custom visual reasoning agents can be deployed on top of existing perception stacks, with claimed annotation budget reductions of 80% or more compared to manual baselines. NomadicML's customers include Zoox, Mitsubishi Electric Automotive America, Natix Network, and Zendar. In March 2026, the company raised an $8.4 million seed round at a $50 million post-money valuation, led by TQ Ventures with participation from Pear VC, Jeff Dean, and angels from OpenAI and Google DeepMind. Key features: - Natural language search over video and sensor data archives - Vision language model pipeline for automated video labeling and event detection - Training dataset curation and edge-case library generation for AV and robotics - Custom visual reasoning agents deployable on existing perception stacks - Fleet monitoring with structured, searchable output from raw footage - Reinforcement learning dataset creation from real-world footage - Supports sensor fusion (video, LiDAR detections) as input evidence
Not public (enterprise sales; pricing page exists at nomadicml.com/pricing but tiers not disclosed in public sources)