At Discover (Google's content recommendation system), I have worked on:
Build revolutionary Discover experiences using GenAI
1. Work closely XFN with multiple teams to figure out product requirement and best solutions
2. Research on LLM solutions to decide on model selection and potential solutions of LLM hallucination
3. Iterate fast on prototype
4. Design and build solution for e2e flow for the experience, with complexity and latency in consideration.
5. Lead multiple rounds of prompt-tuning and live experiments to meet quality bar
Design and build e2e ranking models
1. Roadmap on ranking objective and headroom
2. Build training data pipelines that serve horizontal use for the team
3. Leverage SOTA architecture to build transformer-based ranking models
Build content understanding to surface quality contents