Sieon Lee · AI Engineer · LLM Agents & Retrieval
AI engineer focused on reliable agents, retrieval, memory, and AI infrastructure.
I’m an AI Engineer with a background in data engineering and production systems, completing my M.S. in Applied Data Science at the University of Chicago in 2026.
I focus on LLM agents, RAG, retrieval, vector search, memory, and AI infrastructure. I’m currently looking for full-time AI Engineer roles starting January 2027.
01 Focus
What I work on
LLM agents
I work on agent systems that need to understand user intent, choose and use tools, manage context and memory, and behave reliably across multi-step workflows.
Retrieval and RAG
I’ve worked with embeddings, vector search, metadata filtering, and retrieval pipelines using systems such as OpenSearch and Pinecone.
Production systems
My background is in production engineering, so I tend to think about AI systems beyond the model itself. That includes data pipelines, infrastructure, monitoring, debugging, and what happens when something fails in production.
02 Background
From production systems to AI reliability
I didn’t start my career in AI.
At LG CNS and Nexon, a big part of my job was figuring out why production systems failed. That meant looking through logs, SQL queries, infrastructure metrics, and application behavior to trace problems and prevent them from happening again.
That experience still shapes how I approach AI systems today.
As I started working with LLM agents, I became interested in similar questions. What happens when retrieval gives the model the wrong context? What happens when an agent chooses the wrong action? What happens when old memory conflicts with newer information?
That led me to become especially interested in AI system reliability and failure detection.
03 Research
Current research question
I’m currently exploring whether we can predict when an LLM agent is likely to fail before it generates a response.
My research focuses on cases where an agent has conflicting information in memory, such as an outdated user preference or location, and whether signals available before generation can indicate a higher risk of failure.
Can we detect unreliable AI system states before they turn into incorrect answers or actions?
04 Experience
Relevant experience
Super Amplify — AI & Agent Development Intern
Worked on intent routing, Pinecone-based retrieval, tool integrations, agent workflows, and evaluation for an enterprise AI agent platform.
University of Chicago — M.S. Applied Data Science
Working on a healthcare AI capstone involving retrieval, structured data reasoning, and reliable memory design.
Nexon Korea — Data Engineer
Worked on production data infrastructure, ETL pipelines, AWS, Terraform, monitoring, database performance, and a 500GB+ OpenSearch vector embedding store.
LG CNS — Production systems
Worked on production systems, databases, operational reporting, SQL troubleshooting, and systems reliability.
05 Stack
Technical stack
06 Sieon Labs
What I write about here
Sieon Labs is where I document what I’m building, studying, and trying to understand.
Most of my writing is about LLM agents, RAG and retrieval, agent memory, AI reliability, vector search, and papers related to the problems I’m working on.
Sometimes I write about something I built. Sometimes it’s a failure I ran into. Sometimes it’s a paper that changed how I think about a problem.
07 Contact
Let’s connect
I’m currently based in Chicago and looking for full-time AI Engineer opportunities starting in January 2027. I’m particularly interested in roles involving LLM agents, RAG, retrieval, applied AI, and AI infrastructure.