Principle AI Researcher/Engineer - (LLM reinforcement learning)

Job Description

You join AMD’s Silo AI Base Models team to lead post‑training for large language models. The focus is turning strong base checkpoints into assistant‑grade models using supervised training, reward models, and reinforcement learning at scale, with a mandate to open‑source code, data, and recipes.

Responsibilities

• Design and tune post‑training methods such as SFT, DPO/PPO/GRPO, and related RL variants on large clusters.
• Build high‑throughput synthetic‑data pipelines with clear evaluation metrics.
• Partner with evaluation teams to define metrics that actually track model quality.
• Publish open‑source code, datasets, and training recipes; upstream improvements to TRL and similar frameworks.
• Coordinate with pre‑training, infra, and OpenEuroLLM partners on checkpoints, data mix, and long‑context plans.
• Shape the roadmap to improve multilingual and low‑resource performance in production models.

Benefits