Research Engineer, Gemini Code Post-training, DeepMind

Job description

Drive post-training research and engineering using reinforcement learning (RL) and supervised fine-tuning (SFT) to advance Gemini coding capabilities across web, 3D, game, and mobile development. Develop and scale agentic post-training pipelines and landing recipes in collaboration with Operations Research (OR) teams to establish industry-leading benchmark performance in Code Arena. Design, build, and maintain frontier evaluation suites and automated benchmarks to measure, stress-test, and improve agentic coding capabilities. Implement training infrastructure, reward models, and data curation workflows to accelerate iterative model improvements from revision to revision. Minimum Qualifications: Bachelor's degree in Computer Science, Machine Learning, Artificial Intelligence, or a related technical field, or equivalent practical experience. 5 years of experience in software engineering or research developing machine learning models with frameworks such as JAX, PyTorch, or TensorFlow. Experience conducting post-training, reinforcement learning (RL), or supervised fine-tuning (SFT) for large language models (LLMs). Experience designing or running LLM evaluation benchmarks and measurement pipelines. Preferred Qualifications: Experience in agentic coding workflows, code generation, or software engineering tasks across web, mobile, 3D, or game development. Experience with competitive coding benchmarks, Code Arena, or public model leaderboards. Experience scaling distributed training pipelines on TPU or GPU accelerators. Experience with reward modeling, preference optimization, or synthetic data generation for code models. Experience working in fast-paced research environments delivering iterative model releases.