▸ internship · cohort forming

A few weeks. One benchmark. Real nerds only.

We hire the best — people with outstanding ideas. In 4–8 weeks you'll design and ship a public benchmark for frontier AI models: a capability eval, or a realistic cyber range.

term
4–8 weeks
stipend
$1,200 + $5k API
mode
remote / Warsaw
ship
a public benchmark
apply↓or email internship@arimlabs.ai

01what you get

  • $1,200 stipend + up to $5,000 in frontier-model API tokenspaid
  • A public benchmark with your name on the specyours
  • 1:1 calls with the leadership teammentorship
  • Real fluency in how frontier AI is evaluatedyou'll learn
  • The fastest path to a full-time seat herewhat's next

02who we hire

  • Outstanding ML-research instinct — you read papers and run themresearch
  • Taste for what makes a model capable — and what fakes itjudgment
  • Niche depth in SWE or cybersecurity — CTF, low-level engineering, ML researchcraft
  • Real nerds — the ones who felt the pain during the degree, not just collected itwe mean it

Any rigorous technical background is welcome — physics, applied mathematics, EE — not just CS.

03apply

What you've built or broken, and the benchmark you'd build here. Two minutes.

▸ Your own words only — answers that read as AI slop are rejected immediately.

or email internship@arimlabs.ai