I'm Hongsup/ํ์ญ, an AI/ML engineer with 9+ years of experience spanning production ML systems, agentic AI, and RAG infrastructure. I have a background in computational neuroscience (PhD) and behavioral ecology (MSc). I lead the SciPy Conference Proceedings Committee as co-chair, overseeing peer review of 30+ scientific papers annually, and run a virtual ML journal club. I also volunteer with Texas Justice Initiative, a criminal justice non-profit in Texas.
My work focuses on building reliable AI systems with an evaluation-first approach, from learning-to-rank models for hardware verification to agentic document processing pipelines and enterprise RAG frameworks. I care about making AI systems trustworthy and measurable, whether that means designing ML evaluation frameworks, building retrieval systems, or bridging research and production.
Check out my blog and connect with me on LinkedIn.
Recent work:
- The Corpus-Provenance Gap in RAG Evaluation: Why RAG eval tooling versions everything except the retrieval corpus, and a content hashing pattern to fix it (2026)
- AI SDK Adoption and Conway's Law: Why internal AI platform adoption fails, with practical patterns for user-embedded development (2026)
- Police Data Intelligence Assistant: Multi-agent LangGraph system for automated data enrichment of police shooting records with human-in-the-loop design
Publications & research:
- IEEE SOCC 2024: Robust learning-to-rank algorithm for bug discovery in hardware verification
- IEEE SOCC 2022: Data-centric ML pipeline for hardware verification
- SciPy 2019: ML Ops case study for hardware failure detection
- Officer-involved shootings in Texas 2016โ2019 (with TJI)
- PNAS: Human visual short-term memory (PhD thesis)