Good Start Labs
Research Scientist
$220K–$300K
Remote (US & Canada) · Full-time · Posted Oct 30, 2026
About the job
Build superhuman game-playing agents and use controlled experiments to test and transfer game-based learning into general language models.
What you'll do
- Train and distill expert game agents; design reward signals and training systems.
- Run SFT and RL experiments and develop evaluations for language models and multi-agent games.
- Publish research and feed findings into game environment design.
What they ask for
- Hands-on RL post-training experience with PPO, GRPO, or RLVR.
- Experience designing evaluations for language models.
- Public research record, such as papers, models, or benchmarks used, cited, or built upon.
- Must be located in the US or Canada.
Nice to have
- Built a superhuman game agent, in the spirit of AlphaStar, AlphaGo, or OpenAI Five.
- Loves games and views them as serious training grounds for intelligence.
Pay
$220K-$300K base salary
Tools
PPO, GRPO, RLVR, SFT
AI focus
Train expert game agents and distill their capabilities into general language models; study how game data and game-based RL affect model capabilities and transfer.
Model-written summary · the original posting is authoritative