Arnav Goel

Data science at UC San Diego. I write code and study models, ship products with real users, and help run a jewelry business my family started in 1873.

Python
TensorFlow
Machine Learning
Graph Theory
React
Next.js
TypeScript
SQL
Pandas
PyTorch
NLP
Data Viz
Docker
Node.js
Scikit-learn
Deep Learning
Python
TensorFlow
Machine Learning
Graph Theory
React
Next.js
TypeScript
SQL
Pandas
PyTorch
NLP
Data Viz
Docker
Node.js
Scikit-learn
Deep Learning

About

Data science at the intersection of research and real products

I'm a senior in the BS Data Science program at UC San Diego (minor in Entrepreneurship & Innovation) graduating June 2027 with a 3.911 GPA. My work sits where machine learning meets systems that actually ship — Flask microservices for a 50,000-patient hospital platform at ADA, quantitative research at Triton Quant, and the full-stack platform for Gondilal Saraf, my family's century-old jewelry business.

I care more about a model that's honest about its limitations than one with a flashy accuracy number, and more about code that an on-call engineer can debug at 3 a.m. than code that looks clever in a notebook. Considering new-grad applied-scientist, ML-engineer, and SWE roles for summer 2027 onward — especially in health, commerce, or infra. What I'm looking for →

Selected Outcomes

  • SWE Intern · ADA

    Flask microservices + Nurse Panel API for a multi-region hospital platform projected to onboard 50k+ patients across 120+ hospitals. Cut setup time 60% via Docker.

  • Triton Quantitative Trading · UCSD

    Research & backtest systematic strategies on tick-level data. Feature engineering, time-series normalisation, volatility-adjusted Monte Carlo.

  • Digital Platform Lead · Gondilal Saraf

    Full-stack e-commerce + admin ERP for a 150-year jewelry business. 15 Prisma models, 26 API routes, 85 tests.

  • Power Outages · DSC 80

    Analysed 1,534 U.S. outages (2000–2016). Random Forest with engineered features — R² 0.220, fairness checks across weather vs. non-weather.

Featured Work

Selected Projects

View all

Watch Together — Cross-Site Video Sync

Chrome/Firefox/Safari extension that syncs playback on Netflix, YouTube, Disney+, Prime Video, and more. WebSocket relay with heartbeat drift correction, host mode, ad detection, and built-in chat. 59 server tests.

Chrome ExtensionWebSocketNode.js

Red Bull YouTube Sentiment Analytics

End-term Social Media Analytics project — 500 YouTube comments on Red Bull scored with VADER. Net Sentiment Score of +28.6 pp (double the industry benchmark), 8-chart Excel dashboard, Word report, and executive summary PDF.

PythonpandasVADER

PCOD Tracker — AI Health Companion

AI-powered PCOD/PCOS health app — vent freely and Claude AI auto-extracts symptoms, mood, and meds. Upload lab PDFs for instant hormone parsing. 15 Prisma models, medication streaks, and lab trend charts.

Next.js 16Claude AIPostgreSQL

Gondilal Saraf — Full-Stack Jewelry Platform

Full-stack platform for a century-old family jewelry business — bilingual storefront, live gold rates, AR try-on, Gemini AI descriptions, admin ERP, 15 Prisma models, 26 API routes, and 85 tests.

Next.js 15PostgreSQLGemini AI

U.S. Power Outages — DSC 80

DSC 80 project analysing 1,534 major U.S. power outages (2000–2016). Found higher electricity prices correlate with shorter outages (p ≈ 0.007). Random Forest with log-pop-density + severe-weather features: RMSE 6,189 min, R² 0.220.

Pythonscikit-learnRandom Forest

Buzz — College Event Discovery

Native iOS + macOS event app for US college students, launching at UCSD. Multiplatform SwiftUI (no Catalyst), Supabase with PostGIS + RLS, Stripe Connect ticketing, App Clip, ARKit 'Look Around', and a Next.js 16 PWA mirror.

SwiftUISwift 6Supabase