Arnav Palkhiwala

Arnav Palkhiwala

ML Systems · AI Research · Data Science

Originally from San Diego, I'm a third-year Data Science and Statistics student at the University of Washington (Seattle).

I've spent the past few years building ML systems, focusing on optimizing LLM inference and infrastructure in production. I'm especially drawn to early-stage startups solving real problems.

30%+ Production Throughput Contextual AI
128k Context Window Extension Qualcomm
SIGIR Published '26 UW InfoSeeking Lab
#1 Startup Competition Dempsey @ UW

Work Experience

Atomic Oct 2025 - Present
Venture Open-Source Research Intern
Building an automated ML scouting system analyzing 40k+ repos weekly. Developing an agentic ranking and diligence workflow that combines market indicators, technical feasibility, and developer sentiment to surface commercially viable open-source projects.
Contextual AI June 2025 - Sept 2025
Member of Technical Staff, AI Inference Intern
Led vLLM optimizations improving production throughput by 30%+ on H100/B200 GPUs. Built load-testing and inference-benchmarking systems; metrics informed RAG pricing and helped close new customer contracts.
Qualcomm June 2024 - Sept 2024
Machine Learning Software Engineering Intern
Extended Gemma2 context length from 2k to 128k tokens using Positional Skipwise Fine-tuning (PoSE), boosting long-input accuracy to 90%+ for on-device inference on Snapdragon chips.

Education

University of Washington
B.S. Data Science, Statistics | Minor in Business Entrepreneurship
GPA: 3.85  ·  Dean's List  ·  Expected June 2027

Technical Skills

ML / AI
Python PyTorch Agentic Systems vLLM LangChain RAG Ollama Production LLMInference Benchmarking
Stats
R Statistical Modeling Statistical Inference Gaussian Processes
Web / Infra
TypeScript JavaScript React SQL Docker Kubernetes Firebase Google Cloud HTML/CSS
Other
Java Git Linux Bash Shell

Outside of work I'm into sports, restaurants, and guitar. Always happy to connect!

Reach Out