Arnav Sharma

CS @ Penn State | Intern @ IBM

~/about

I work on AI engineering, LLM inference, model serving, evaluation, and reliable data systems.

I'm a Computer Science student at Penn State. Outside of work, you'll find me hiking, cooking, or trying new foods.

~/what i've done

~/projects

CacheForge

2-bit KV-cache quantization for LLM inference with custom CUDA kernels; 2.6x lower peak memory and up to 3.47x higher throughput.

pythonpytorchcudac++flashattention

Benchseed

Parallel, constant-memory TPC-H data generation in Rust with Arrow and Parquet; byte-for-byte parity in CI and 14.4x faster than DuckDB at 100 GB.

rustarrowparquetpythoncli

Threadwave

Pure-Go real-time collaboration with Yjs-compatible CRDTs, WebSockets, SQLite, and NATS/JetStream, verified across 158 cross-language fixtures.

gocrdtswebsocketssqlitenats

Automated Predictive Modeling

full-stack ML platform with real-time inference and explainability

nextjspythonfastapi

ASL Learning Platform

98.98% accuracy model with real-time hand tracking in-browser

pytorchmediapipenextjs

PSU Dining AI

AI-powered dining menu nutritional analysis

pythonflaskgemini api

Resume Chatbot

RAG-powered chatbot for conversations about my background

pythonlangchainopenai

Reddit Post Analyzer

AI classification and summarization for financial subreddit posts

pythongemini apihttpx

~/contact

<aqs7726@psu.edu>

or say hi on linkedin