Data Science & Economics · UC Berkeley

I’m Tristan.

I am a designer, researcher, student, teacher.

Selected work

Experience

  1. Sep 2026 – Present

    Archaeological Research Facility, UC BerkeleyBerkeley, CA

    Undergraduate Research Assistant

    • Built the BD3D research viewer (Next.js, TypeScript, React Three Fiber), linking 314 scholarly annotations to 607 surface regions across 6 photogrammetric Egyptian coffin models, with two-way selection, raking-light inspection, and collection-wide search.
    • Designed a compressed 3D asset pipeline (meshopt glTF, tiered textures, checksummed provenance) that cut exhibit downloads by 74% (82 MB to 21 MB across six coffins) and geometry parsing from 82–132 ms to 2–3 ms.
    • Built a structured annotation data pipeline with schema validation, replacing manual XML lookup; 127 automated tests run in CI.
    • Lead the incoming URAP students and the design team.
    Read the case study →
  2. Sep 2026 – Present

    Ten StrandsBerkeley, CA

    Data Research Intern

    • Built an NLP retrieval system combining BM25, dense transformer embeddings, and a sentence-transformers CrossEncoder reranker over 10,000+ California K–12 district webpages and policy documents.
    • Benchmarked lexical and neural retrieval on human-validated evidence (Recall@K, MRR): Recall@5 up 35%, manual source-discovery time down 76%.
    Read the case study →
  3. May 2026 – Aug 2026

    IDX ExchangeRemote

    Data Science Intern

    • Built a model that estimates California single-family sale prices from 309K cleaned CRMLS sales (Jan 2024 – Apr 2026), adding location, property-age, and school-district features via a spatial join across 680 districts.
    • Evaluated chronologically with leakage rules: the final XGBoost model reached 7.7% median error on 11,890 held-out April 2026 sales, versus 16.5% for a linear baseline.
    Read the case study →
  4. Aug 2025 – May 2026

    Lucarelli Photogrammetry Lab, UC BerkeleyBerkeley, CA

    Undergraduate Research Apprentice

    • Built an automated Python photogrammetry pipeline (Agisoft Metashape API) processing 1,000+ images into 3D reconstructions, cutting manual processing time by 40% (~30 → ~18 min each).
    • Developed Ridge regression models on SfM and image-overlap graph features to predict reconstruction quality: 13.7% lower geometric RMSE than a reprojection-error baseline, with leakage-aware cross-validation.
    Read the case study →
  5. Aug 2025 – PresentJan 2026 – May 2026

    Berkeley ANova

    Computer Science Mentor

    • Teach Python weekly to 100+ high school students.

    Marketing Officer

  6. Jan 2025 – May 2025

    Goldman Sachs

    Possibilities Summit

  7. Aug 2024 – Feb 2025

    Rose Water CapitalUnited States

    Business Consulting Intern

    Design and Advertising

  8. Feb 2023 – May 2024

    Shaolin Temple Cultural Center

    Martial Arts Coach

Education & skills

University of California, Berkeley

B.A. Data Science & Economics

Expected May 2029 · GPA 3.88

CourseworkMicroeconomics, Probability and Statistics for Business, Foundations of Data Science, Computational Structures in Data Science, Linear Algebra

Languages

  • Python
  • SQL
  • JavaScript
  • TypeScript
  • HTML/CSS

Frameworks & tools

  • React
  • Next.js
  • FastAPI
  • PostgreSQL
  • Supabase
  • MongoDB
  • Git
  • Vercel
  • OAuth 2.0
  • REST APIs

Data science & ML

  • scikit-learn
  • PyTorch
  • Sentence Transformers
  • OpenCV
  • Pandas
  • NumPy
  • SciPy
  • Matplotlib
  • Tableau

About

I study Data Science and Economics at UC Berkeley (class of 2029). I like problems where a model or a tool is only half the job and the other half is making it useful to the people who rely on it: researchers, teachers, students.

Every week I teach Python to 100+ high schoolers as a Computer Science Mentor at Berkeley ANova. Explaining an idea to a room of fifteen-year-olds is the fastest way I know to find out whether I really understand it.