Thomas Liu

Math of Computation @ UCLA

I build machine learning models and full-stack tools, from glucose prediction with wearables to research-triage pipelines at Berkeley Lab.

Experience

Lawrence Berkeley National Laboratory

Jun – Jul 2025

Data Science Intern

Berkeley, CA

  • Built an internal tool that filters hundreds of research papers for alignment with LBNL's 2024–2034 strategic goals, cutting literature triage from a month-long process to minutes.
  • Curated a dataset of academic papers and fine-tuned Gemini 2.5 Pro, at no additional cost to the lab, to power the paper-relevance pipeline.
  • Adopted across departments; collaborated through Bitbucket version control.

Projects

TouchBase

2026

Next.js · TypeScript · PostgreSQL · Supabase · Claude

  • Full-stack app that helps students find and cold-email professors.
  • Data pipeline pulling researcher and publication data from OpenAlex and web search, with Claude summarizing each professor's research focus across 140 professors.
  • Recommendation engine suggesting similar professors by scoring research-topic overlap.

Callsense

2025

3rd place overall · Milpitas Hacks

Next.js · TailwindCSS · Flask · Gemini

  • 911 operator interface that records, transcribes, and live-summarizes calls to reduce operator burnout.
  • Placed third out of 150 competitors.

Non-invasive Glucose Prediction with Wrist-Worn Wearables

2025

Bio-Rad Laboratories award

PyTorch · LSTM · GRU

  • Collected Fitbit data and trained ML/DL models to predict glucose levels non-invasively.
  • Ensemble LSTM + GRU model predicted 98.8% of values within a clinically acceptable range.

Predicting Diabetes from Lifestyle Factors

2024

AP Research

MLP · TabTransformer · scikit-learn

  • Built ML/DL models that improved accuracy, sensitivity, and specificity over existing methods by ~15%.
  • Presented at a research symposium; recognized as one of the year's best projects.

Education

University of California, Los Angeles

2026 – 2030

B.S. Mathematics of Computation

Differential Equations, CS 31

Diablo Valley College

2024 – 2026

Concurrent enrollment · GPA 4.0

Linear Algebra, Multivariable Calculus, Discrete Math

Monte Vista High School

2022 – 2026

GPA 4.5 / 3.9

National Merit Finalist Scholarship Recipient · AIME Qualifier · 3× hackathon winner

Leadership

Stang Hacks

Jan – Apr 2026

Director of Outreach

Dublin, CA

  • Drove 450+ sign-ups for one of the largest high school hackathons in the Bay Area.
  • Recruited and managed 20+ volunteers for day-of operations.
  • Partnered with 13 sponsors (Notion, Vercel, ElevenLabs, Cerebras, Omi, and more) for $4k+ in prizes.

Skills

Languages
Python, TypeScript, JavaScript, Java, CSS
ML & Data
PyTorch, scikit-learn, Pandas, NumPy, Matplotlib
Web & Tools
React, Next.js, TailwindCSS, PostgreSQL, Supabase, Git