This role focuses on advancing the evaluation and development of frontier coding agents, positioned at the intersection of AI research, software engineering, and model evaluation. You will design the benchmarks, methodologies, and data systems that shape how next-generation coding models are measured and improved.
Key Responsibilities - Design and own evaluation frameworks for coding agents, including benchmark specifications, scoring methodologies, rubrics, and quality standards.
- Lead end-to-end research initiatives focused on measuring and improving coding model performance across diverse software engineering tasks.
- Develop high-quality datasets, golden examples, and evaluation protocols that enable reliable assessment of frontier coding systems.
- Analyze model behavior and failure modes, identifying systematic weaknesses and translating findings into actionable improvements for training and evaluation.
- Build tooling an...