Frontier Benchmarking Research Engineer

Refresh is seeking a Research Engineer focused on benchmarking, to build rigorous evaluations of real-world coding and computer-use capabilities. You will translate expert workflows into verifiable tests and run benchmarks against frontier models.

In this role you’ll work across our core stack (Vercel, Supabase, Render) and engage with reinforcement learning and supervised fine-tuning at a high level to publish robust numbers.