Research Engineer / Scientist, Post-Training

What you'll do

• Train models for better agentic tool use, especially context management, and develop post-deployment weight updates that avoid catastrophic forgetting.
• Design experiments on data mixtures, training algorithms, and models; build infrastructure for large-scale synthetic data generation and collection.
• Create evaluations for agentic capabilities and publish research through papers, technical reports, blog posts, and open-source code.

What they ask for

• Proficiency in Python and deep learning frameworks such as PyTorch.
• Expertise in post-training methods, such as SFT fine-tuning, reinforcement learning, reward models, or preference learning.
• Ability to balance execution speed with empirical rigor.
• Proven impactful research track record and real-world impact beyond academic work.

Tools

Python, PyTorch

AI focus

Post-training LLMs for agentic tool use and context management; continual post-deployment learning without catastrophic forgetting; synthetic data and agentic evaluations.