Kaiwei Luo
Software engineer based in San Francisco. I build event-driven backend systems and the AI features that sit on top of them.
export const kaiwei = { role: "Software Engineer", company: "Amazon", location: "San Francisco, CA", focus: [ "event-driven architecture", "GenAI / LLM integration", "platform migration", ], education: "UMD, B.S. CS + Math", openTo: "interesting problems",};
About
README.mdI've been at Amazon since 2022, where my work spans two layers. On the backend, I move high-volume systems onto event-driven architecture, most recently a seller compliance scoring system that now makes enforcement decisions in minutes instead of hours.
Closer to customers, I build AI-powered features: a search widget that hands conversational queries to an AI assistant, retrieval sources for an LLM ads system, and recommendations on voice-assistant smart displays. I also wrote an MCP server of agent scripts that cut our on-call troubleshooting time by two-thirds.
Before Amazon I did support engineering at L-Soft and integration work at fusionSpan. I studied Computer Science and Mathematics at the University of Maryland.
Experience
git log --oneline-
Aug 2022 – now
● currentSoftware Development Engineer II
Amazon · San Francisco, CA
- Seller compliance scoring migration. Led the end-to-end move from legacy infrastructure to an event-driven architecture: real-time event routing, rule-engine integration, shadow and load testing, and fault-tolerant failover. Processing latency went from 18h → 3m, enabling near real-time enforcement for millions of marketplace sellers at lower infrastructure cost.
- AI assistant entry point in search. Designed and built a search widget that detects conversational queries and surfaces AI assistant capabilities, covering query classification, product context retrieval, dynamic UI, state management, and handoff to chat, with zero impact on search-page latency.
- MCP server for on-call. Built an MCP server hosting AI-agent scripts for teammates on call, cutting troubleshooting time across six services from ~15m → ~5m and automating daily checks of deployment pipeline health and ticket queues.
- LLM ads evidence retrieval. Developed a sponsored brand summary generator used as a retrieval source for Amazon's post-editing LLM ads system, improving the relevance and accuracy of ads shown to customers.
- Smart display recommendations. Built personalized product recommendations from purchase history for Android-based voice-assistant smart displays, end to end: UI, product categorization, schema changes, and recommendation API integration.
-
Nov 2021 – Aug 2022
Software Support Engineer
L-Soft · Bethesda, MD
- Primary support engineer for on-premises and cloud clients, owning tickets through to accurate, thorough resolution.
- Set up new hosting clients and configured services to their requirements.
-
May 2021 – Oct 2021
Java Developer
fusionSpan · Rockville, MD
- Built a data pipeline migrating e-commerce business data into Salesforce CRM with TIBCO Cloud Integration.
- Wrote Python scripts connecting REST APIs to integrate internal company data.
Skills
stack.json{ "languages": ["TypeScript", "JavaScript", "Java", "Python", "Kotlin", "C#", "SQL", "HTML/CSS"], "frontend": ["React", "Web Components", "Responsive Design", "Design Systems"], "backend": ["Event-Driven Architecture", "Distributed Systems", "REST APIs", "Microservices"], "aws": ["Lambda", "DynamoDB", "S3", "SQS", "CDK", "CloudWatch", "Bedrock", "IAM"], "tools": ["Git", "Unix/Linux", "CI/CD"]}
What I'm good at
- System design
- Platform migration
- GenAI / LLM integration
- A/B testing
- Shadow & load testing
- Agent tooling (MCP)
Earlier projects
Coursework and side projects from college, from data science notebooks to small web apps, are on my GitHub.
Education
University of Maryland, College Park
B.S., double major in Computer Science and Mathematics
Major GPA 3.58 / 4.00
Contact
The fastest way to reach me is email. I'm also on LinkedIn and GitHub.