Improving Gemini models for agentic capabilities such as Deep Research at Google DeepMind
Requirements
Experience working on software engineering projects from proof-of-concept through to implementation.
Experience productionizing state-of-the-art large language and multimodal models, prompt optimization.
Knowledge of LLMs, machine learning and statistics
Experience in applying experimental ideas to applied problems
Required programming languages: Python; C++
Experience post training large models (e.g. SFT, RL)
Responsibilities
Designing Intelligent Agents: Create and implement sophisticated agentic workflows that allow our models to intelligently use tools and data to solve complex problems.
Pragmatic System Design: Make critical design trade-offs based on a deep understanding of model capabilities and limitations. You'll architect systems that are not just theoretically sound but are also cost-effective, reliable, and delightful for users at scale.
Driving Model Improvement: Develop robust evaluation systems, autorating mechanisms, and synthetic data pipelines to ensure our models are constantly learning and improving.
End-to-End Feature Ownership: Shepherd new agent capabilities from initial ideation and prompt engineering, through system design and implementation, to launch and post-launch analysis. You will be the expert on the why behind a feature, not just the how.
Pushing the Boundaries of LLMs: Experiment with novel post-training techniques to unlock new capabilities and efficiencies in our models.
Elevating the Team: Mentor and guide fellow team members, contributing to a culture of technical excellence and innovation.
Other
BS, MS or PhD degree in computer science, mathematics, applied stats, machine learning or similar experience working in industry
Great communication skills and interpersonal skills
Ability to work in ambiguity and have a shipping mindset, willing to help out with whatever moves prototypes forward.
Flexibility and adaptability to work on any project is a must
Background check performed by a third party acting on our behalf
