Harim Choi

Machine Learning Engineer / Researcher

Sejong Bid Institute • Temps plein

Mar 2026 - Aug 2026 • 5 mos

Developed and operated probabilistic machine learning and decision-support systems for public procurement. Built regional tabular models for bidder counts and winning-price distributions, and combined uncertainty estimates with competitor-behavior modeling and Monte Carlo simulation. Compared models and policies on time-based, out-of-sample evaluations. Improved the bid-award KPI by 30% over the previous single-quantile model and reduced processing time from 136 seconds to 25 seconds. Managed data preparation, training, batch inference and deployment, with versioned artifacts and reproducible runtime contracts.

Machine Learning Engineer / Researcher

Hanmac Group

Nov 2024 - Feb 2026 • 1 yr 3 mos

Built machine learning and automation systems for public procurement across Hanmac Group. Developed bid-eligibility classifiers using weak supervision and RoBERTa-large with LoRA, achieving Macro F1 0.9639. Deployed INT8 ONNX models through FastAPI on CPU at approximately 50 ms per notice in the measured environment. Built probabilistic bid models and Monte Carlo decision systems, improving the group-wide PQ bid KPI by 35%. Developed LangGraph RAG workflows connecting document retrieval, cost estimation, quantity takeoffs and draft generation.

Data Analyst

Nielsen

Aug 2019 - Oct 2020 • 1 yr 2 mos

Worked in Nielsen Korea GTC on global retail product-image and barcode-data workflows. Used computer vision and robotic process automation to automate product-image analysis, attribute extraction and barcode-database construction. Built exception-handling workflows so only cases requiring manual review were routed to analysts, reducing department processing time by 40%.