Experience
Experience
Engineering Lead/Manager, ElevenLabs (August 2025 – Present)
Engineering lead and product owner for ElevenAgents, the company's enterprise vertical agents platform. Lead a team of backend, frontend, and product engineers across 3 countries.
Senior Manager (Head of AI), Amazon Customer Service (March 2022 – August 2025)
Founded and scaled a 50-person applied research and ML engineering team building AI for Chat, Voice, and Search across 2B+ annual customer interactions.
Led the org's generative AI strategy from prototype in 2023 to deployment across 7 countries, 6 languages, and 5 business units (Associate operations, Retail, Alexa, Prime Video, Drivers, Amazon Business) by 2025. Improved CSAT by 500 bps, automated 100M contacts end-to-end, drove ~$300M in 2024 cost savings, and reduced average handling time by ~12 seconds across NA and UK support associates.
Architected a hybrid agent framework combining structured workflows with autonomous LLM behavior, balancing reliability in critical sub-flows with flexibility for long-tail queries. ~20 agent applications launched on this framework; showcased by Andy Jassy at AWS re:Invent 2025.
Built pre-launch testing, red-teaming, and real-time policy guardrails for production agent deployment. Mentored 10 research interns and published in retrieval, LLM uncertainty, evaluation bias, post-training, reasoning, multilingual adaptation, and RL-based routing.
Tech Lead Manager, Amazon Physical Stores (October 2020 – March 2022)
Founding member of Amazon Style, an AI-driven physical fashion store in LA combining in-store fitting with online-scale selection through interactive fitting rooms and computer vision.
Architected and deployed a real-time multimodal recommendation system using online learning, boosting purchase rate by 2000 bps over baseline. Scaled tech hiring from 5 to 20 and led the public launch.
Senior Applied Scientist, Prime Wardrobe (April 2018 – October 2020)
ML lead for Amazon Try Before You Buy. Launched a human-in-the-loop recommendation system with neural recommenders and multi-arm bandits, improving Try Rate by ~800 bps over human stylists. Built outfit-building algorithms; proposed and shipped PSANet and OutfitTransformer.
Applied Scientist, AWS SageMaker (February 2017 – April 2018)
Contributed to the AWS SageMaker and AWS Comprehend launches at re:Invent 2017. Developed SpectralLDA, a tensor-decomposition-based topic model with ~4X speedup over Mallet.
Data Scientist, Mu Sigma (May 2013 – July 2015)
Built flight-delay simulation models for Southwest Airlines, ~97%-accuracy laptop sales forecasts for Dell, and Bayesian Marketing Mix Models attributing revenue to marketing spend.