I'm curious by nature, and I happen to spend my days in the DV (design verification) world — 10+ years across the front-end verification flow for CPU and GPU silicon, writing verification plans and building the C++ golden reference models our sign-off flow leans on, and more recently verifying and enhancing the infrastructure around that flow: simulation tooling, coverage automation, and the debug tooling verification teams reach for daily without thinking twice about it.
That same curiosity is what got me building with LLMs — wondering if a model could take the tedious parts off my plate, like chasing down why tests failed, triaging logs, suggesting the fix before I've had my coffee. It turned into a habit: whenever I hit a problem worth automating, at work or at home, I build something to solve it and see how far the model can take me.
This profile is where that curiosity turns into shipped code — my playground for AI projects. Some solve problems I run into in daily life (family finances, retirement planning), some are me kicking the tires on a new model just to see what it can do, and all of them start the same way: I hit something annoying, and figured I'd try building my way out of it.
- 🔭 Building AI tools to solve problems I actually run into
- 🧠 Currently deep in where LLM agents meet hardware verification — RTL-aware debuggers, automated triage, tool assistant
- 🎓 M.S. Computer Engineering, Texas A&M University
- 🏆 1st place, Hack@DAC (DAC 2018)
| Project | Description | Tech Stack | Live |
|---|---|---|---|
| 🔧 RTLGraph | Built on top of the TPE accelerator RTL as a testbed — a semantic retrieval engine that turns a compiler's elaborated view of a hardware design into a queryable graph AI agents can reason over, instead of retrieving RTL source text. Graph traversal only: no embeddings, no vector search, no LLM in the retrieval path. | Python · FastAPI · NetworkX · SQLite · React · TypeScript · Cytoscape.js | Live ↗ |
| 🔍 rtl-lens | Built to evaluate how well a graph database can power real RTL debugging — an AI agent that diagnoses failing hardware simulation tests by querying RTLGraph (fanin/fanout, drivers, clock/reset domains) instead of embeddings, proposes and verifies fixes through a human-approval checkpoint, and grades its own diagnoses against a hidden answer key. Fixed 4 of 6 targeted bugs blind on the TPE accelerator testbed. | Python · LangGraph · LangChain · LangMem · Anthropic Claude · Streamlit | Live ↗ |
| 🔧 tpe-tensor-processing-engine | Built as the base framework I'll develop several AI automations on top of — a from-scratch AI accelerator with RTL, a C++ golden model, and a full pyuvm/cocotb verification environment, including 7 intentionally-catalogued bugs to exercise the flow. |
RTL · C++ · UVM · Verification Methodology | Live ↗ |
| Project | Description | Tech Stack | Live |
|---|---|---|---|
| 🤖 agentic_ai_basics | My cheatsheet for agentic AI concepts — a learning series of small, self-contained Python templates for building with the Claude API. | Python · Claude API · Agentic Patterns | Live ↗ |
| Project | Description | Tech Stack | Live |
|---|---|---|---|
| ⚽ meridian-strike | A fun project inspired by the recent FIFA World Cup — a browser-based 3D soccer game with full 11v11 matches, tiered AI opponents, and a career ladder. | JavaScript · Three.js · Node.js | Live ↗ |
| 🎮 road-rage | Built to put the Fable5 model's capabilities to the test — a browser-based synthwave motorcycle combat racer in Canvas 2D / vanilla JS, single HTML file. | HTML5 · Canvas · JavaScript | Live ↗ |
| Project | Description | Tech Stack | Live |
|---|---|---|---|
| 🌱 nourish-habit | Built to help me prepare healthier meals without giving up the dishes I already eat, so I can stay healthy without a diet-food overhaul — a free meal-planning PWA that upgrades your habitual meals to hit a calorie target, cooks from your own pantry, and finds recipes through an AI chat. | React · TypeScript · Vite · Tailwind CSS · Dexie/IndexedDB · PWA | Live ↗ |
| Project | Description | Tech Stack | Live |
|---|---|---|---|
| 💰 familyfinance | A personal app for managing family finances and planning for retirement — self-hosted dashboard with Plaid account aggregation, expense tracking, net worth/property tracking, reports, and a Monte Carlo retirement planner. | Python · Plaid API · Monte Carlo Simulation | — |
| 💸 expense-tracker-ai | My first project, built just for fun with Claude Code — a private, local-first personal expense tracker. | Next.js · TypeScript · Tailwind CSS | Live ↗ |
