Skip to content
View lyhjeremy's full-sized avatar

Block or report lyhjeremy

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
lyhjeremy/README.md

Hi, I'm Jeremy Lee 👋

Data scientist and UCLA Anderson MSBA. Formerly Chief Strategy Officer; now focused on applied analytics and machine learning. I turn ambiguous business problems into models and decision tools.

npx lyhjeremy

↑ run that in your terminal for the interactive version of this README

What I work with

  • Modeling · scikit-learn · PyTorch · XGBoost · regression / classification / clustering · time series · neural networks · causal inference · survival analysis · A/B testing
  • Optimization · Gurobi · LP / IP / QP / non-convex programming · branch-and-bound · LP duality · gradient descent · simulation · Monte Carlo
  • GenAI · LLM prompting · RAG · embeddings · vector databases · fine-tuning · OpenAI / Anthropic APIs · NLP
  • Agents & orchestration · LangChain · LangGraph · agentic workflows · tool use · retrieval pipelines · multi-step reasoning
  • Data & tools · Python · R · SQL · Snowflake · Airflow · Spark / PySpark · Tableau · Power BI · Excel · Git · GCS · Jupyter
  • Deployment & apps · GitHub Pages · Streamlit · interactive dashboards · desktop apps · live demos · reproducible pipelines

Featured projects

Selected work spanning generative AI, statistical modeling, and machine learning. Each title links to the code; the Live link opens a plain-language write-up in the browser. The full set of projects is in the dropdown below.

Project Methods
Wine Sommelier RAG Semantic search over 130k wine reviews using hybrid BM25 and dense retrieval with a cross-encoder reranker, evaluated on retrieval quality (precision and nDCG up roughly 21%). Live
Marathon Training Agent A LangGraph agent that iteratively drafts and revises a training plan until it satisfies explicit safety checks, with a Streamlit interface. Live
World-Record Half-Lives Survival analysis (Kaplan-Meier and Cox regression) of how long track-and-field world records stand, estimating each record's probability of falling. Live
Performance Continuum Distribution fitting and extreme-value theory locating any marathon time within a field of 31,842 finishers and quantifying how far the world record sits beyond it. Live
National Strength Profiles K-means clustering and PCA of 43 nations across five athletics disciplines, identifying five distinct performance archetypes. Live
Scoring-Table Fairness Audit Non-parametric hypothesis testing (Kruskal-Wallis) on 6,400 elite marks, assessing whether World Athletics points are comparable across events. Live
More projects (16 additional repositories)

Generative AI & agents

Project Methods
Marathon Strategy RAG Grounded retrieval-augmented Q&A that cites its sources, evaluated with an LLM-as-judge faithfulness metric; includes a LangChain LCEL implementation. Live
Wine Pairing Agent A LangGraph agent that infers a wine style and calls the retriever as a LangChain tool, with full run tracing. Live

Running & sports analytics

Project Methods
The Heat Tax Regression of Boston Marathon field times on temperature: roughly one minute slower per °F (r = 0.86). Live
The Negative-Split Myth Split analysis showing only 2.5% of 31,912 Boston finishers ran a faster second half. Live
Hitting the Wall Per-5K split analysis locating where marathon pace decays, around 30 km. Live
Boston BQ Fairness An equity analysis of Boston qualifying standards across age and gender brackets. Live
Marathon Majors Course Difficulty A three-framework comparison of finishing speed across the six World Marathon Majors. Live
Marathon Shoe Revolution Decomposition of recent marathon progression into footwear effects versus other factors. Live

Language & NLP

Project Methods
Are Song Lyrics Getting Simpler? Fifty years of Billboard lyrics: longer and more repetitive over time, with roughly stable vocabulary. Live
Love Island Lexical Analysis A corpus study of lexical complexity in reality-television dialogue. Live
EPUB Reading-Level Analyzer An in-browser tool reporting a book's reading level and vocabulary load from an EPUB file. Live
Mandarin Learning Reader An interactive Mandarin reader built from any Chinese EPUB, with per-character pinyin, read-aloud, and offline pronunciation grading. Live
Cantonese Learner A mobile Cantonese speaking tutor built from current Hong Kong news articles. Live

Consumer & finance

Project Methods
Wine Score Inflation How 130k professional reviews cluster into an effective 8-point band on the nominal 100-point scale. Live
When Can I Retire? A Monte-Carlo retirement simulator estimating the probability a portfolio lasts through retirement. Live

Apps & tools

Project Methods
Leave Time Optimizer A routing tool that computes when to depart to arrive on time, using the Google Maps API. Live
EPUB → Audiobook A desktop app that converts e-books into read-along audiobooks (English and Chinese). Live
Podcastify Converts an e-book or article into a two-host podcast with synthetic voices. Live

Experience & education

  • Co-founder & Chief Strategy Officer, Casual Ace Learning Centre — grew enrollment 5.2× to over 1,000 students across six centers.
  • MSBA, UCLA Anderson · graduate analytics coursework, Georgia Tech · BBA (Accounting & Finance), University of Hong Kong.

Outside the terminal

14 marathons, Berlin personal best of 2:48 · WSET Level 3 in wine.

Contact

LinkedIn · Email

Pinned Loading

  1. epub-to-audiobook epub-to-audiobook Public

    Turn any EPUB into a natural-sounding audiobook on your own computer, with read-along highlighting. Free, open-source, English + Mandarin + Cantonese.

    Python

  2. boston-marathon-qualifying-fairness boston-marathon-qualifying-fairness Public

    Are Boston Marathon Qualifying Times Fair?

    HTML

  3. leave-time-optimizer leave-time-optimizer Public

    Google Maps Leave Time Optimizer - know exactly when to set off!

    HTML

  4. marathon-majors-course-difficulty marathon-majors-course-difficulty Public

    Are the World Marathon Majors Equally Fast? A Three-Framework Analysis of Course Difficulty Across the Six Majors

    HTML

  5. cantonese-learner cantonese-learner Public

    Cantonese Learner v2 (mobile) — speak Cantonese from today's Hong Kong news, iPhone-friendly

    JavaScript

  6. marathon-strategy-rag marathon-strategy-rag Public

    A retrieval-augmented marathon coach that cites its sources — grounded in real marathon science and my own analyses of 30k+ Boston finishers.

    Python