Instance Analysis
What is Instance?
Instance provides a verification layer for robotics that returns a success verdict with evidence for every robot episode using camera footage. The company's fine-tuned model judges each attempt pass/fail, breaks videos into grounded subtasks with live captions, and outperforms Claude Opus 4.8 on held-out test sets. Founded by MIT computer scientists with backgrounds at SpaceX, AWS, and NASA JPL and backed by Y Combinator, Instance addresses the data bottleneck in robot learning.
Location
San Francisco, United States
Employees
1-10
Founded
2026
Industry
B2B
AI Pattern
AI infrastructure / tooling
Builds Hardware
Software only
Business Model v4
Developer infrastructure
Delivery Model
Developer platform / API / infra
Extruct Category
Hardware
Founders LinkedIn
- Claire Mao: https://www.linkedin.com/in/clairemao-
- Lucy Cai: https://www.linkedin.com/in/lucyjcai
Sub-cluster
robotics verification
YC Partner
Ankit Gupta
Product Description
Instance offers a verification layer for robotics that provides a success verdict (pass/fail) for each robot episode, supported by evidence derived from camera footage. The main product features include:
- Verification API: This API returns a clear success verdict for each robot episode, allowing users to assess the performance of their robots effectively.
- Fine-tuned Model: The model is designed to evaluate each attempt, categorizing it as a pass or fail based on the evidence provided.
- Grounded Subtasks: The system breaks down videos into manageable subtasks, each with live captions, enhancing the clarity and usability of the data.
- High Performance: Instance's model has been shown to outperform Claude Opus 4.8 on held-out test sets, indicating superior accuracy and reliability.
Benefits:
- Data-Driven Insights: Users receive detailed insights into robot performance, helping to identify areas for improvement.
- Efficiency in Robot Learning: By addressing the data bottleneck in robot learning, Instance enables faster and more effective training of robotic systems.
- Targeted for Professionals: The product is specifically designed for robotics engineers, robot learning researchers, and AI/ML teams, making it a specialized tool for those in the field.
Overall, Instance's offerings are aimed at enhancing the capabilities of robotics through precise verification and analysis.
See something that needs updating? Suggest edits to this profile.