Someone who wants to build out what LLM systems can really do
Bunjang is Korea's leading secondhand marketplace, with 10M+ monthly active users. I've spent five years there building production ML systems on AWS, and recently focused on stateless, high-throughput LLM inference optimized for cost and latency at scale. I'm now looking to build out a wider range of what LLM systems can do — agentic systems in particular.
Projects
In-Chat Abuse Detection System
I helped build a system that detects escrow-payment-evasion abuse in Bungaejangter's in-chat feature using an open-source LLM. It sanctioned 50% more abusive users than the legacy system, driving a shift to secure payment and growing escrow transaction volume — this walks through how it was built.
AI-Assisted Product Registration System
To ease the manual burden of filling out required fields on the product listing screen, I built an assistant feature where, depending on the field, an ML model or an LLM generates and recommends a value. It increased the listing conversion rate by 4pp and listings per user by 47% — this walks through how the system was built.
Similar Product Recommendation System
I helped build a system that surfaces products similar to one a user just clicked, in real time, across multiple recommendation surfaces. It increased total clicks in the home feed's and product-detail page's recommendation areas by 13% and 9% respectively — this walks through how it was built.