Developing the next generation of AI technologies to empower Google users, specifically working on Project Janus, a new product development in Gemini App team
Requirements
- Have hands-on experience on LLM quality iteration
- Have experience with building real AI systems, especially actual agents for products
- Have a strong learning mindset, and eager to learn and apply new technologies to real work
- Are excited by and experienced with going 'zero to one' - creating something new, starting from a set of broad ideas in a small team environment
- Are highly cross-functional and able to work collaboratively across different PAs
- Excel at breaking down abstract Critical User Journeys into actionable ML tasks, quickly creating the right evaluation and SFT datasets to improve the model
- Bring a deep curiosity for model behavior, excelling at turning detailed analysis of outputs into targeted data and quality enhancements
Responsibilities
- Develop from scratch the next generation of AI technologies that empower Google users
- Work on Janus modeling and quality, including setting up evaluation set and infrastructure, agentic orchestration flow, prompt iteration, context engineering, autorater, training data collection, working with One Recipe team on iterating on new version of Gemini model through SFT / RL techniques
- Regularly read research and industrial ideas in this area and apply to our real use case
- Proactively participate in code reviews, enforce development best practices, and share your knowledge to raise the team's overall technical expertise
- Work in a fast paced small 'whatever it takes' team environment to rapidly develop a new product
- Excel at breaking down abstract Critical User Journeys into actionable ML tasks, quickly creating the right evaluation and SFT datasets to improve the model
- Bring a deep curiosity for model behavior, excelling at turning detailed analysis of outputs into targeted data and quality enhancements
Other
- In office role for MTV, with no consideration for additional sites at this time
- Must be able to work in a fast-paced small team environment
- Must be willing to 'roll up your sleeves' and 'do the work' within a small-team environment
- Must be able to work collaboratively across different PAs
- Must have a strong learning mindset and be eager to learn and apply new technologies to real work
