Hemut / Agents and backend systems
A logistics copilot that finds the tools a question needs and coordinates answers across them.
Find every part
Select relevant tools
Organization-scoped
Grounded in results
I built the agent layer for questions that needed several answers from different tools. The existing guardrails got in the way of complex, multi-part queries. Retrieving tools for the whole message could miss a secondary part of the question, so the work centered on covering each part and coordinating the calls needed to answer it.
Embedding-based retrieval selected relevant tools from the wider catalog. My backend work also moved database access onto organization-scoped sessions enforced by PostgreSQL row-level security, then addressed the query plans behind a slow handler.
The reviewed baseline has 75 passing prompts out of 229, or 32.8%. My resume reports an end-to-end accuracy improvement from 33% to 99%. That is a separate reported result, not a demonstrated follow-up evaluation of those same 229 prompts.
Original resume outcome for the MCP agent layer and embedding-based tool retrieval. The reviewed baseline is recorded separately; no paired after-evaluation artifact is supplied for the 99% figure.
75 passing prompts divided by 229 reviewed prompts, rounded to one decimal place. This supports the baseline only, not a subsequent accuracy result.
Tool catalog size reported in the original resume, with embedding-based retrieval selecting relevant tools. This is an implementation-scale figure, not an accuracy measurement or a current registry audit.
Original resume reports organization-scoped, row-level-security-enforced sessions across a nine-organization Postgres estate, followed by query-plan fixes. The reported handler timings have no specified percentile or load.
Agent and backend engineering by DongYeop Lee at Hemut.
Voice agents for freight calls, with work spanning the conversation, the backend, and how both reach production.
A logistics copilot that finds the tools a question needs and coordinates answers across them.
A speech-recognition benchmark built around freight calls, including a scoring correction that changed my model comparison.
