Vision Lab
Category: Robotics / Embodied AI
Industrial data layer for robotics training — capturing and structuring real factory workflows at scale to train robotics and AI models. Vision Lab was founded in 2025. The company is led by Tanachart (James) Kujareevanich. Based in San Francisco, California, United States. Team size: 11-50. Total funding raised: $6M. Latest round: Seed. Key investors include Y Combinator, Undisclosed Silicon Valley investors.
- Founded
- 2025
- Headquarters
- San Francisco, California, United States
- Team size
- 11-50
- Total funding
- $6M
Value proposition
Vision Lab provides the missing data infrastructure for physical AI by capturing real-world industrial workflows from 2,000+ factories across 27 countries, pairing first-person video with SOP-level process knowledge to train robotics foundation models on authentic production environments rather than lab demos.
Products and solutions
Project Watermelon (atomic action with clip-level task diversity, 1,000+ environments), Project Strawberry (frame-perfect dense annotation and hand pose extraction), Project Cherry (atomic action annotations with expert validation), Project Pear (industrial data SOP capture with engineer validation), Proprietary temporal VLM for industrial settings, High-density action labeling with proprietary VLMs and human-in-the-loop
Unique value
The only company building a structured industrial dataset at scale from real factory environments (not lab simulations), with a proprietary temporal VLM that outperforms Gemini by 2.5x on industrial temporal understanding, serving 4 of the top frontier AI labs.
Target customer
Frontier AI labs, robotics companies, and foundation model builders needing real-world industrial training data; manufacturing partners who contribute data for licensing revenue.
Industries served
Manufacturing, Automotive, Electronics, Semiconductors, Biotech/Pharma, Medical Devices, EV, Chemicals, Furniture, Jewelry, Textiles, Food Processing
Technology advantage
Proprietary temporal VLM trained on SOP-paired industrial footage that outperforms Gemini by 2.5x on temporal understanding in industrial settings; Global factory capture network with standardized protocols, verified ground truth, and clean data rights; 2,000+ factories across 50+ industries in 27 countries; First-person egocentric industrial video with SOP-level process knowledge
How they differentiate
Unlike robotics foundation model companies that build end-to-end models trained on lab data, Vision Lab focuses exclusively on the data bottleneck — capturing real factory workflows at scale with SOP-level annotations. Their proprietary temporal VLM for industrial settings outperforms Gemini by 2.5x. They have the most diverse industrial dataset spanning 50+ industries across 27 countries with 2,000+ partner factories.
Main competitors
Physical Intelligence, Skild AI, Covariant (via Amazon), Config (Korea-based, described as "TSMC of robot data")
Key partnerships
Y Combinator (Spring 2025 batch), 2,000+ partner factories across Asia and Africa (primarily China and India), 4 of the top frontier AI labs as clients (including 3 of the Magnificent Seven tech companies)
Notable customers
4 of the top frontier AI labs (client identities withheld under NDA), 3 of the Magnificent Seven technology companies
Major milestones
Founded 2025, Accepted into Y Combinator Spring 2025 (X25) batch, Built network of 2,000+ factories across 27 countries, Raised $6M in seed funding (June 2026), Serving 4 of the top frontier AI labs including 3 Magnificent Seven companies, Partner factories earned $1M+ in data licensing revenue, Developed proprietary temporal VLM outperforming Gemini 2.5x on industrial tasks
Market positioning
Vision Lab positions itself as the data infrastructure layer for physical AI, distinct from robotics foundation model builders (Physical Intelligence, Skild AI, Covariant) by focusing exclusively on the data supply side — capturing real-world industrial training data rather than building robot brains. Competes indirectly with Config (robot data platform) but differentiates through global factory network scale (2,000+ factories) and proprietary industrial temporal VLM.
Geographic focus
Global (HQ in San Francisco, factory network across Asia and Africa primarily in China, India, and expanding into Southeast Asia)
About Tanachart (James) Kujareevanich
Ex-McKinsey Ops Consultant; MBA from MIT Sloan. Former McKinsey operations consultant before founding Vision Lab.
Official website: https://thevisionlab.ai