Research Engineer/Scientist

Lead research on LLM mid-training and post-training methods such as continued pretraining, supervised fine-tuning, preference optimization, and reinforcement learning, driving experimental and data decisions that impact downstream agent behavior.

Innovate and prototype new agent architectures and algorithms focusing on components like planning, reasoning, memory, skills, tool use, retrieval, and multi-agent collaboration beyond current state-of-the-art approaches.

Develop and own research methodologies including evaluation benchmarks for reasoning, planning, reliability, and safety; identify failure patterns in models and translate insights into research or training improvements.

3+ years of applied machine learning, deep learning, or AI research experience with independent project ownership.

Hands-on experience with LLM training/post-training approaches including at least one of continued pretraining, supervised fine-tuning, preference optimization, or reinforcement learning.

Strong Python and advanced PyTorch skills, including modifying models and training pipelines for experimentation.

Experience designing evaluation methodologies and benchmarks, including trajectory-based and human evaluations.

Experienced in agentic AI topics such as planning, reasoning, memory, tool use, retrieval, and long-context processing with knowledge grounding.

Proven ability to independently identify research problems, generate hypotheses, conduct rigorous experiments, and drive projects to measurable impact.

Skilled at communicating complex research findings effectively to technical and executive audiences and collaborating to integrate research into production.