Andy Li
Backend and AI engineer with 6+ years building and running LLM-powered systems, and the cloud infrastructure behind them, on AWS.
I take LLM features from idea to production: RAG grounded in real data, tool-calling agents that check with a person before acting, all kept validated, cached, and within budget.
About
Shipping LLM features responsibly, on infrastructure I can stand behind.
Backend and AI engineer with 6+ years building and running LLM-powered systems and the cloud infrastructure behind them on AWS. I'm comfortable taking an LLM feature from idea to production and just as comfortable owning the infrastructure it runs on.
Opinionated about shipping AI responsibly: a person signs off on sensitive actions, answers stay grounded in real data, model output gets validated, and someone keeps an eye on cost.
On the AI side, I wire up OpenAI or Anthropic, add RAG search over real data using embeddings and pgvector with citations back to the source, and build tool-calling agents in Python and FastAPI that look things up and draft work but hand off to a person before doing anything sensitive. I keep the whole thing validated with JSON schema checks, cached, rate-limited, and within budget.
On the infrastructure side, I deploy and operate services on AWS ECS with Terraform, integrating DynamoDB, S3, and Kinesis, and standardize rollout and rollback across the team. I design real-time backend services in Node.js and TypeScript with bounded timeouts, caching, and Redis-backed idempotency to hold p99 latency and strict availability targets.
I instrument services with metrics, structured logs, and CloudWatch alerts to shorten detection and rollback during incidents, and I carry on-call ownership, usually leaving latency and uptime better than I found them.
Professional Experience
Building and operating scalable backend systems with a focus on performance, reliability, and maintainability.
Senior AI & Cloud Engineer
Yelp
- Built a caching and rate-limiting layer in front of OpenAI and Anthropic calls, with JSON schema checks that retry or fall back on bad model output, cutting cost and tail latency and keeping malformed responses away from users
- Built retrieval in Python over our own content using embeddings, pgvector, chunking, and citations back to the source, grounding answers in real data and reducing hallucinations on customer-facing questions
- Built Python tool-calling agents in FastAPI that sort requests, look things up, and draft replies but hand off to a person before anything sensitive, with an audit log and per-request cost tracking
- Designed and operated real-time backend services in Node.js and TypeScript with bounded timeouts and caching, improving p99 latency and supporting strict availability targets on high-traffic consumer APIs
- Added Redis-backed idempotency for retries plus metrics, structured logs, and CloudWatch alerts, reducing duplicate processing and cutting detection and rollback time during incidents
- Deployed services on AWS ECS with Terraform and integrated DynamoDB, S3, and Kinesis, standardizing rollout and rollback across the team
Cloud Software Engineer
Best Buy
- Improved release safety with CI/CD enhancements, staged rollouts, and rollback-ready Docker deployments on AWS, shortening deploy cycles
- Developed Node.js and SQL-based services for commerce and marketplace workflows that stayed stable under high-volume retail traffic and peak events
- Optimized PostgreSQL and MySQL queries and service paths to maintain stable response times under load
- Collaborated with DevOps on Docker-based deployments and monitoring, improving detection and deployment consistency
- Introduced integration tests and pre-deploy checks for critical paths, reducing production defects and speeding up safe releases
Full Stack Software Engineer
Open Systems International
- Built backend APIs in Node.js, REST, and SQL for real-time monitoring and control systems, enabling predictable real-time behavior
- Designed clear API contracts with explicit timeout and edge-case handling, reducing production surprises from downstream failures
- Supported on-call, operational debugging, and incident mitigation on Linux-based environments, improving response time
- Automated build and deploy steps and documented runbooks, cutting deploy time and rollout errors
- Added tests and refactored hot paths for critical workflows, and mentored junior engineers on backend reliability and system design
Notable Projects
Key projects showcasing scalable backend architecture and performance optimization.
LLM Cost & Reliability Layer
Built a caching and rate-limiting layer in front of OpenAI and Anthropic calls, with JSON schema validation that retries or falls back when the model returns bad JSON. Wrapped the whole path in metrics and per-request cost tracking.
Impact: Cost and tail latency dropped, and malformed responses stopped reaching users.
Grounded RAG Search
Built retrieval in Python over our own content using embeddings, pgvector, chunking, and citations back to the source, so answers could be traced and checked instead of taken on faith from the model.
Impact: Responses got grounded in real data and hallucinations fell on customer-facing questions.
Human-in-the-Loop Agent Platform
Built Python tool-calling agents in FastAPI that sort requests, look things up, and draft replies, but hand off to a person before doing anything sensitive, with an audit log and per-request cost tracking for every action.
Impact: Cut a lot of manual work without giving up control or auditability.
Resilient Real-Time Backend
Designed and operated real-time backend services in Node.js and TypeScript with bounded timeouts, caching, and Redis-backed idempotency, deployed on AWS ECS with Terraform alongside DynamoDB, S3, and Kinesis.
Impact: Improved p99 latency, met strict availability targets, and cut duplicate processing and rollback time during incidents.
Core Skills
Technologies and practices I use to build scalable, reliable systems.
AI & LLM Engineering
Languages
Backend & Systems
Cloud & Infrastructure
Data & Caching
Reliability & Practices
Education
Formal grounding in computer science fundamentals.
Bachelor of Science in Computer Science
University of Minnesota, Twin Cities
Get in Touch
Based in Plymouth, MN. Open to senior AI and backend engineering roles. Let's connect.