Key Features
Engineering Details
Stack
CS student at Penn State who builds things end-to-end. Shipped RAG systems to production, led a team, won a capstone. Looking for SWE or ML roles starting May 2026.
Senior at Penn State, graduated May 2026 with a 3.86 GPA and 8 consecutive semesters on the Dean's List. Major in Computer Science.
I've built RAG systems used by 200+ Penn State employees, shipped a capstone app now adopted by engineering firms, and written everything from PyTorch transformers to C memory allocators from scratch.
Led 6 interns building Drawing Vision, a CV system for engineering drawing metadata. Architected WorkPro, a production RAG system cutting query latency 70% for 200+ Penn State OPP users on Azure Container Apps with Azure AI Search.
Built a RAG chatbot automating government RFP analysis, cutting 3+ hours of manual review per submission. Built reusable AI tooling and Medicaid/Medicare data dashboards for executive reporting.
In-person and remote support for 500+ students and faculty. Authored user documentation adopted across 200+ community members.
Production-ready full-stack app automating dimensional tolerance calculations for aluminum extrusions per industry standards (Tables 11.2 & 12.2). Won Penn State Engineering Capstone against 76 teams — now actively used by aluminum associations and engineering firms.
Browser extension + financial dashboard recommending optimal credit cards per purchase, flagging scam sites via IPQS, and analyzing spending with Gemini RAG. Stores user data in MongoDB Atlas with personalized income/savings reports.
Drag-and-drop neural network builder that compiles graphs to real PyTorch and trains on Modal cloud GPUs (A10/T4) with live loss streaming. Supports 30+ layer types, deterministic shape inference, and a Claude-powered graph-aware AI tutor.
Smart calendar that aggregates tasks from Canvas, Outlook, and Jira into a unified task lake, auto-schedules them into free slots, and launches context-aware AI agents when a task block starts — opening relevant files and a focused assistant.
Tournament harness that pits five LLM system-prompt archetypes against each other in attacker/defender matchups, with a separate judge model scoring every exchange. Produces a dominance matrix to test whether any prompting style is universally robust or a rock-paper-scissors structure emerges.
Custom liballocator in C implementing Buddy and Slab allocation with automatic coalescing, header-based tracking, and sorted free lists. Validated against 500+ allocation/deallocation test cases with diff-based output matching.
Decoder-only transformer built from scratch in PyTorch — multi-head self-attention (4 heads), positional embeddings, layer norm, residual connections. Trained 4000 epochs on character-level text, capturing meaningful linguistic patterns.
Real-time surgical instrument tracking system built to reduce retained-tool incidents in operating rooms. Uses OpenCV and PyTorch classifiers to monitor tool state across three categories — in field, in use, and discarded — with a live activity log for post-operation analysis. React frontend displays current tool status; Flask backend handles detection and state management.
Looking for full-time SWE or ML engineering roles from May 2026. I require H-1B sponsorship.