Mechanize
Reinforcement learning environments for AI labs
- 1-20 employees
- B2B
- Artificial Intelligence
- Machine Learning
- SoMa, San Francisco, CA
Jump to section
Company mission
To develop virtual work environments, benchmarks, and training data that will enable the full automation of the economy.
Our take
Mechanize develops Reinforcement Learning environments that AI labs can use to train their models. The company believes that the greatest value of AI will be in automating ordinary tasks. However, current models lack the long-context capabilities and multimodality to execute long-term plans, which is what Mechanize's approach hopes to address.
Basically, the company supplies an environment in which AI models can be assigned tasks and graded on how well they complete them. These tasks can include things like building features, deploying apps or debugging issues. Then, the grade for the task asks as a reward signal to help the AI learn when it has done well vs badly.
In supplying these RL environments, Mechanize supports the more effective automation of everyday tasks, enabling businesses to extract real value from AI. The company is backed by investments from Nat Friedman and Daniel Gross, Patrick Collison, Dwarkesh Patel, Sholto Douglas and Marcus Abramovitch.
Kirsty
Company Specialist at Welcome to the Jungle
Leadership
Matthew Barnett
(Co-Founder)Formerly Staff Researcher at Epoch AI, Script Writer at Rational Animations and Data Scientist at Metaculus.
Tamay Besiroglu
(Co-Founder & CEO)Previously co-founded Epoch AI, was a Research Scientist at MIT and Strategy and Operations Lead at Metaculus.
Ege Erdil
(Co-Founder)Worked alongside Matthew and Tamay at Epoch AI.