Job Summary
The role involves overseeing the complete post-training processes for agentic models, focusing on everything from SFT warm starts to extensive on-policy reinforcement learning (RL) sessions. The position aims to enhance the overall performance, reliability, and efficiency of the models while guaranteeing that the training operations can be replicated across different computing environments.
Responsibilities
- Lead the full-spectrum post-training activities for agentic models.
- Manage SFT warm starts and large-scale on-policy RL executions.
- Optimize model performance, reliability, and efficiency.
- Ensure reproducibility of training runs in various compute environments.
- Monitor and adjust training parameters to achieve desired outcomes.
Qualifications
- Proven hands-on experience in training or fine-tuning large language models (LLMs) using RL techniques.
- Familiarity with distributed training frameworks, particularly PyTorch.
- Strong experimental discipline and methodological approach to research.
- Ability to collaborate effectively and work within European Union regulations.
