At Amazon, I architect Alexa+ systems for LLM routing, retrieval-augmented generation, and capability awareness. I remodeled the LLM-facing API and evals, cutting routing anomalies from 60% to 1%.
I drove delivery of an automation pipeline that learns Alexa+'s capabilities from API specs and device UI screenshots, supporting discovery and proactive recommendations. I also re-architected personality retrieval, improving accuracy by 90% on personality customer requests.
Earlier at Amazon, I delivered the AWS GameSparks JavaScript container runtime and optimized its OSGi module runtime, cutting AWS Lambda cold starts from over 10 seconds to under one. On Kindle Digital Pricing, I architected workflows that cut pricing re-drive time from 3 hours to 10 minutes.
