Member of Technical Staff, Applied Research

The role

You will design and run model improvement workflows including supervised fine-tuning, preference optimization, and reinforcement learning. You will build environments, tasks, reward functions, and training infrastructure for frontier AI research.

What you will do

• Design and run model improvement workflows across supervised fine-tuning, preference optimization, and reinforcement learning.
• Design tasks, tools, validators, reward signals, and evaluation harnesses.
• Create, inspect, and improve data for teaching models how to behave.
• Work on reward functions, rubrics, validators, and analysis tools.
• Run experiments across distributed GPU clusters using PyTorch and FSDP.
• Build infrastructure for model training, evaluation, and data generation at scale.
• Translate ambiguous research goals from frontier AI labs into concrete deliverables.

Required qualifications

• Masters or PhD-level knowledge of machine learning fundamentals.
• Understanding of linear algebra, optimization, and stochastic gradient descent.
• Experience with or deep understanding of LLM fine-tuning, preference data, reward modeling, or reinforcement learning for language models.
• Fluency in Python and PyTorch.
• Comfortable building with FastAPI and modern ML infrastructure.

Nice to have

• Roughly 3-5 years of experience

About the company

Chakra Labs builds high-fidelity environments, evals, and datasets for frontier AI research. The team consists of ex-Stripe, Snap, AWS, Microsoft, and Airtable members.