LLM-powered search
Designed and shipped a search tool over sensitive databases, reducing analyst lookup time from minutes to seconds. Built the retrieval pipeline, prompt architecture, and backend API from scratch.
Mark Liudmirsky
01 / BUILD
Four years of building intelligent tools and operational software. From retrieval pipelines to systems used in the field.
2022–2026. Built and shipped software across AI-powered search, visitor management, and operational event replay. Owned retrieval pipelines, backend APIs, data models, and application logic.
Designed and shipped a search tool over sensitive databases, reducing analyst lookup time from minutes to seconds. Built the retrieval pipeline, prompt architecture, and backend API from scratch.
Built a desktop management platform that replaced a fragmented Excel workflow, cutting manual entry from 30 minutes to seconds per visitor. Designed the data schema, access-control model, and REST APIs serving entire battalions.
Worked on a C# replay engine, including a data-ingestion pipeline, state-reconstruction logic, and a 3D rendering layer for operational sensor data.
An AI-powered market summarization service built on a microservices architecture.
View mktsum on GitHub ↗A chat-based interface for connecting to SQLite databases and executing queries through Discord.
View SQLite Discord Bot on GitHub ↗A browser-based toolbox for encoding, hashing, string and entropy analysis, hex inspection, and disassembly.
View Web RE Toolbox on GitHub ↗