Building safety and trust infrastructure for Microsoft Copilot - consumer and enterprise AI experiences across M365 core.
Key accomplishments
Built LLM-judge evals (intention accuracy, takeover quality) to measure and validate consent policy compliance across Cowork MSA flows for OneCopilot surfaces; authored the quality-metrics-model spec defining canonical KPI taxonomy, metric owners, and quality thresholds.
Shipped a 30+ tile Azure Data Explorer dashboard (KQL, ADX) with unit-tested metric definitions and live health/risk reporting, driving cross-surface consent decision quality for Copilot Spear Security.
click to flip back
Software Engineer
Amazon | Alexa Autonomous Agent (AAA)
Seattle, WA • May 2025 – June 2026
Multi-agent AI at Amazon scale
click to flip
Software Engineer
Amazon | Alexa Autonomous Agent (AAA)
Seattle, WAMay 2025 – June 2026
Building Alexa+, Amazon's next-gen conversational AI with autonomous agent orchestration and multi-agent task execution.
Key accomplishments
Building pipeline infrastructure for Alexa+: multi-agent orchestration and autonomous task execution at scale for Amazon's next-gen conversational AI.
Developing LearningCompanion content pipeline: deep research, content planning, and quality validation in a multi-agent Python/Java framework; automated personalized text digests (scheduling & notifications).
Integrated MCP adapter connecting Python-based LLM agents to Java backend services; transport API, integration tests, cloud coverage reporting, and prompt testing for orchestration workflows.
Implemented performance state management with MemoryDB: persistence for complex task dependency graphs; DSAR compliance workflows for agent memory (enterprise data governance & privacy).
click to flip back
Software Engineer
Amazon | Alexa Central Analytics & Research Science
Bellevue, WA • June 2024 – May 2025
10× cheaper, millions of devices
click to flip
Software Engineer
Amazon | Alexa Central Analytics & Research Science
Bellevue, WAJune 2024 – May 2025
Integrated caching and data-sharing solutions for Alexa+ pipeline; shipped notification system for BanyanLite device.
Key accomplishments
Integrated PDM caching microservice into the Alexa+ orchestration pipeline, implementing cache-hit/miss routing that bypasses content sourcing and ranking services on cache hits, reducing backend calls and latency for millions of Echo Show devices.
Shipped notification system for BanyanLite, Amazon's new Alexa device launched Dec 2024, collaborating cross-team to deliver on a tight hardware launch deadline.
Implemented SQS batch API integration consolidating individual message sends into batched operations, achieving 10X cost reduction in queue processing.
Built ShareLink microservice enabling secure one-click transfer of 2000+ utterance IDs between Impact 2.0 and Sleuth platforms, eliminating manual copy/paste and reducing data leakage risk for data scientists.