Jan 2020 – Present · 6 yrs 2 mos · San Francisco Bay Area

• Leading a product team to build GenAI Modeling and Runtime Platform that powers the Amazon's assistants, like Rufus and Alexa+. My team works with 200+ engineers and scientists to solve unique challenges in the LLM stack. We build specialized LLMs for voice and conversational use cases CPT, SFT or RL to serve domain specific purposes. We have shipped products to millions of customers that provide accurate, reliable answers by using grounded, real-world data to minimize hallucination. We also developed a sophisticated orchestration system that selects the optimal model for any task, balancing capability, cost, and latency to operate efficiently at Amazon's scale.
• Also previously:
• Led Alexa Third-Party Skill ecosystem by building AI products and developer tooling; shipped successful products like Name-Free Interaction (for small developers) and Custom Agents (for Enterprises incl. Disney and BMW)
• Led Multimodal initiatives to reimagine CX for voice-assistants with screens; built proactive intelligence to leverage visual screen space to offer relevant recommendations, and developed self-learning AI solutions to automatically recover from defective customer interactions.