Savit Pawar

Personal record · updated 2026

Savit Pawar

Role
CS researcher & student
Location
Seattle, Washington · 47.66°N 122.31°W
Studying
B.S. Computer Science, University of Washington
GPA
3.91 / 4.0
Status
Open to internships & research
Focus
ML systems · agentic tooling · full-stack
See selected work Résumé (PDF) 

About

Hello, and welcome. I'm a computer science student at the University of Washington, where I also do research in the Programming Languages & Software Engineering lab and TA a systems course.

I work across the stack, lately on agentic LLM tooling, ML systems, and full-stack products. I'm drawn to problems where careful engineering makes something measurably better, and most of what I learn ends up as a project. A few are logged below.

Outside of computer science I follow Formula 1 and football closely.

Experience

  1. Jun 2026 – Present

    Undergraduate Researcher

    University of Washington · PLSE Lab

    Building an agentic LLM pipeline that translates C into byte-accurate, test-verified Rust, with a reproducible Google Test harness for equivalence checks. Extended an LLVM call-graph tool to catch indirect function-pointer call sites the standard pass misses.

  2. Jun 2026 – Present

    Teaching Assistant

    University of Washington · Computer Science

    Lead weekly quiz sections and discussions for a 40-student systems course, grade labs and quizzes, and answer questions on the course forum.

  3. Nov 2025 – Jun 2026

    Lab Systems Intern

    University of Washington · SEAL Lab

    Led a Claude Code chatbot build that cut per-person weekly reporting from 45 to 10 minutes across a 50-person lab, and maintained two lab-wide tracking systems for a measured 37% efficiency gain.

  4. Sep 2025 – Present

    B.S. Computer Science

    University of Washington · 3.91 GPA

    Coursework: Data Structures & Parallelism, Hardware/Software Interface, Discrete Mathematics, Linear Algebra, Calculus 3, Intro to Software Engineering.

Selected work

Projects I've built or contributed to.

  1. Sokoban Environment

    Result
    1st in track · 2nd overall
    Stack
    Python · JavaScript · Three.js · BFS
    Role
    Contributor

    A reinforcement-learning benchmark that tests how well LLM agents perceive 3D space, with elevation and box-pushing mechanics, BFS-verified solvable maps, and a 2D/3D browser viewer that replays an agent's decisions move by move. Won its track and placed second overall at a Google-sponsored, Google-judged hackathon.

  2. Talus

    Result
    0.90 AUC-ROC
    Stack
    Python · XGBoost · scikit-learn · SHAP · Earth Engine
    Role
    Contributor

    A statewide landslide-susceptibility model for Washington at 10 m resolution. XGBoost with spatial cross-validation over 450M pixels and 8 features from 7 geospatial sources; on a blind holdout of the 2014 Oso zone it flagged 46.8% of landslide pixels as high risk against 15% of stable terrain. SHAP ranked precipitation and elevation as the top predictors.

  3. LaunchHQ

    Result
    40+ users · 20+ live waitlists
    Stack
    Next.js · Vercel · Neon · Stripe · Resend
    Role
    Creator

    A full-stack SaaS platform for launch waitlists, with an embeddable signup widget, a referral system, custom domains, and Stripe-metered billing. Grew to 40+ users and 20+ live waitlists, and ranked launchhq.space onto Google's first page through hands-on SEO.

Résumé

The tools I reach for, and the full one-page document.

Languages
Java · Python · JavaScript · SQL · Rust · C · HTML/CSS
Frameworks
React · Next.js · Tailwind CSS · Streamlit
AI / ML
XGBoost · scikit-learn · SHAP · pandas · NumPy · Hugging Face · function calling · prompt engineering
Tools
Git · Copilot · Claude Code · VS Code · Google Test · Power BI
PawarSavit.pdf PDF · opens in a new tab
View résumé  Download

Contact

The fastest way to reach me is email. I read everything.