
Created by an AI Research Team from Meta, Microsoft, SoftBank, Tesla, Stanford University, Tsinghua University, and the National University of Singapore.
Runs independently for multiple weeks on a large multi-node GPU cluster, conducting research with web search, code execution, and additional proprietary tooling to power multi-agent swarms.

Takes ownership of entire codebases for weeks at a time. Plans and executes multi-sprint roadmaps, builds new features, writes tests, triages bugs, and ships to production with no human in the loop.

Runs multi-week long training campaigns across GPU clusters. Writes and optimizes CUDA kernels, sweeps hyperparameters, analyzes loss curves, and reruns experiments until convergence.

Whether you're interested in joining the team or exploring an enterprise partnership, we'd love to hear from you.
Get in Touch