# Ryan Reddy > Developer Portfolio. Computer science student at Purdue who builds AI products and the systems behind them: a solo knowledge-graph platform for company documents, on-device vision-language model benchmarking on Snapdragon phones, and a state-winning autonomous robot. - Email: [ryangreddy@gmail.com](mailto:ryangreddy@gmail.com) - Phone: 512-766-9833 - GitHub: - LinkedIn: - Resume: [PDF](/resume-with-camp.pdf) ## Projects Things I have built: AI products, benchmarking tools and a competition robot. ### PhoneBench _Aug 2026 – Present_ A website for queueing benchmark jobs on a shared phone. Teammates queue jobs on the team's Snapdragon phone from a website instead of using USB and adb. Live since October 2026, with sign-in limited to team members. **Highlights** - Kotlin worker app runs each job on the phone's CPU, GPU or NPU - FastAPI and Supabase queue with Google sign-in; React site on Vercel - First cloud job completed all 100 runs (10 images × 10 repetitions) **Tech:** Kotlin, FastAPI, Supabase, React, Vercel **Links:** [Live site (team only)](https://lpcvc-phone-bench-app-server.vercel.app) ### On-device VLM Benchmarks _Aug 2026 – Present_ Qwen3-VL-2B speed on a phone's CPU, GPU and NPU. For the Purdue LPCVC team: a benchmark harness on Qualcomm's GenieX Android sample that records time-to-first-token, prefill speed and decode speed on each processor. **Highlights** - GPU and NPU cut time-to-first-token 2–3× versus the CPU over 10 runs per configuration (cold start: 30.1 s to 9.6 s) - Accuracy pilot scoring real-vs.-tampered image verdicts against a labeled dataset **Tech:** Qwen3-VL-2B, Qualcomm GenieX, Android / adb ### CompanyBrain _Jul 2026 – Present_ A solo-built knowledge graph for company documents. Turns company documents into a source-linked knowledge graph for AI agents. Claude extracts each fact with its source and authority (policy or tribal knowledge). **Highlights** - A measured 0.75 merge threshold replaced a guessed 0.85 that missed a duplicate - Semantic search ranks entities by pgvector similarity, verified on live data - LLM spend capped by a ledger and a fail-closed budget gate **Tech:** Next.js, TypeScript, Prisma, PostgreSQL, pgvector, Anthropic API, Voyage AI **Links:** [Demo](https://company-brain-nu-puce.vercel.app) ### Robot Tour Autonomous Robot _2022 – 2024_ 1st place at the Texas State Tournament. A Science Olympiad robot that has to drive a fixed course the same way every run. I was lead architect; my partner led the code development. **Highlights** - C++/Arduino motion control with encoder distance tracking and IMU/gyro yaw estimation - PID-style correction, S-curve acceleration, drift correction, turn calibration - Custom PCB with gyroscope and Hall effect encoder feedback **Tech:** C++, Arduino, PID control, IMU / gyro, PCB design ## Experience Where I have worked, led and studied. ### Purdue VIP LPCVC Team · Low-Power Computer Vision Challenge _Aug 2026 – Present_ On-device vision-language model benchmarking and remote-access infrastructure. - Built PhoneBench and the Qwen3-VL-2B benchmark harness (see Projects) ### B.S. in Computer Science · Purdue University _Aug 2025 – May 2029_ Expected graduation May 2029. - GPA: 3.44/4.0 ### Program Activity Leader · Children's Association for Maximum Potential (CAMP) _Summer 2025_ A full-time summer camp for children and adults with disabilities. - Led daily recreational and social activities, with hundreds of campers per session across 6 sessions - Adapted each activity to individual campers' needs ## Skills Languages, frameworks and tools I use. ### Languages - Python - TypeScript - JavaScript - Kotlin - C++ - C# - Java - SQL - HTML - CSS - Bash ### Web & Backend - React - Next.js - Node.js - FastAPI - Prisma - PostgreSQL / pgvector - Supabase - Neon - Vercel - Tailwind CSS - REST APIs - pytest - Git ### AI & Data - Anthropic API (tool use) - Voyage AI - Embeddings and vector search - Knowledge graphs - TensorFlow - matplotlib - Claude Code - OpenAI Codex ### Systems & Hardware - Android / adb - Arduino - Unity - Logisim - NgSpice ## About & Contact Who I am and how to reach me. ### Who I am I'm a computer science student at Purdue (B.S., expected May 2029) who builds AI products and the systems behind them. I'm most interested in source-grounded agents, coding tools, and the places AI meets robotics and infrastructure. I move quickly, then stay with the details that make a system useful and reliable. ### Awards & honors Competition results, all from 2024. - Top 30 USA, PicoCTF - AIME Qualifier - CyberPatriot Nationals Semifinalist - 1st, Texas State Tournament: Science Olympiad Robot Tour - 2nd, Texas State Tournament: Science Olympiad Solar Power - FBLA Nationals Qualifier ### Get in touch - Email: [ryangreddy@gmail.com](mailto:ryangreddy@gmail.com) - Phone: 512-766-9833 - GitHub: - LinkedIn: - Resume: [PDF](/resume-with-camp.pdf)