Post-Training Research Engineer, Coding AI

Google; Mountain View (CA); On-site; USD 174,000 - 252,000; Full time

Job summary

Google DeepMind is seeking a Research Engineer focused on Gemini post-training to advance coding capabilities through RL and supervised fine-tuning. You will collaborate with Operations Research teams to build scalable post-training pipelines and benchmark suites that push the limits of agentic coding across domains.

The role emphasizes rapid experimentation, collaboration with academic partners, and contributing to high-impact research and product innovations for billions of users.

Qualifications

• Bachelor's degree in CS/ML/AI or equivalent practical experience.
• 5 years of software engineering or research experience in ML model development with JAX/PyTorch/TensorFlow.
• Experience with post-training RL or SFT for large language models.
• Experience designing or running LLM evaluation benchmarks and measurement pipelines.

Responsibilities

• Drive post-training research and engineering using RL and SFT to advance Gemini coding capabilities across web, 3D, game, and mobile development.
• Develop and scale post-training pipelines and landing recipes with OR teams to establish benchmark performance in Code Arena.
• Design, build, and maintain frontier evaluation suites and automated benchmarks to measure and improve agentic coding capabilities.
• Implement training infrastructure, reward models, and data curation workflows to accelerate iterative model improvements.

Skills

ML model development

RL/SFT for LLMs

LLM evaluation benchmarks

Education

Bachelor's degree in CS/ML/AI or equivalent

Tools

JAX

PyTorch

TensorFlow

Job description

Google DeepMind is seeking a Research Engineer focused on Gemini post-training to advance coding capabilities through RL and supervised fine-tuning. You will collaborate with Operations Research teams to build scalable post-training pipelines and benchmark suites that push the limits of agentic coding across domains.

The role emphasizes rapid experimentation, collaboration with academic partners, and contributing to high-impact research and product innovations for billions of users.