📍 Central Florida · 🤖 Forward-Deployed AI & Systems Engineer · 🎓 CS @ UCF (Dec 2026)
I design, deploy, and operate autonomous AI systems and production SaaS — multi-agent automation, media pipelines, and trading bots on one side; WebRTC voice platforms, power dialers, and CRMs on the other. I ship fast, automate everything, and let the work speak for itself.
Architected a real-time sports prediction market: edge-device camera feeds from Ray-Ban Meta glasses capture the live game, Gemini orchestration extracts team stats on the fly, and the system streams back dynamic win probabilities as the play unfolds.
Gemini LLM Orchestration Ray-Ban Meta Real-Time Vision Prediction Markets
Contributing fixes to the data & AI infrastructure I build on every day — 1 merged upstream, more in review across major projects:
| Project | Contribution | Status |
|---|---|---|
| streamlit | Fix RuntimeError: reentrant call inside _io.BufferedWriter on Ctrl+C |
✅ Merged |
| pandas | Fix loc/iloc setitem with empty / boolean column indexer on EA dtypes |
🟢 Open |
| scikit-learn | Fix OrdinalEncoder crash on numeric transform data with object-dtype categories |
🟢 Open |
| dask | Skip rechunk in to_zarr when chunks already align with the target array |
🟢 Open |
| Chart.js | Don't run tooltip content callbacks when all items are filtered out | 🟢 Open |
| cypress | Don't inject into commented-out head/body/html tags |
🟢 Open |
🎙️ Meeting Copilot — Real-time meeting & interview copilot Open Source
Listens to any call, waits for the real question, and answers from your own knowledge base — 100% local and private.
Captures system audio, transcribes on-device with Whisper (GPU), detects when a speaker actually finishes talking (a wall-clock turn detector, so it ignores rambling and filler), then streams a first-person answer grounded in your own notes from any OpenAI-compatible LLM. Bring your own API key; your notes never leave your machine.
Electron Node.js faster-whisper OpenAI-compatible LLM System-audio loopback
🎬 ClipPro — Free Opus Clip Alternative Open Source
Turn long videos into viral clips — runs locally, no subscriptions, no cloud uploads.
Built with Whisper (GPU) + Gemini AI + FFmpeg. Transcribes your video, scores the highest-virality moments with AI hook ranking, then cuts 9:16 clips with karaoke captions and hook overlays. Includes a web UI and a clip combiner.
Python Flask faster-whisper Gemini 2.5 Flash FFmpeg OpenCV
🖼️ Meta AI MCP — Free image + video generation via Meta AI Source Available
Generate AI videos from Claude Code for free — no API key, no subscription.
Reverse-engineered Meta AI Vibes to expose a working MCP tool and CLI. Browser-driven CDP automation because Meta sends prompts over protobuf WebSocket — no pure-HTTP client exists. Text→video and image→video both work. ~30-90s per clip.
Python MCP Chrome CDP agent-browser FFmpeg Reverse Engineering
🔗 LinkedIn MCP — LinkedIn Browser Automation MCP Source Available
Post, comment, like, and DM on LinkedIn from Claude Code — no official API, no $300/month tools.
Browser-driven CDP automation using your real logged-in Chrome session. Five MCP tools: post (with image), comment, like, DM, and session status check.
Python MCP Chrome CDP LinkedIn
🎯 no-backtest-drift — Live/Backtest Parity Pattern Source Available
Stop your trading bot from lying to you. One pattern: live engine delegates to backtest engine. Zero drift.
The #1 silent killer in algo bots is live-vs-backtest divergence. This pattern fixes it structurally — the live engine has no entry logic of its own; it calls the backtest engine's exact should_enter(). One brain, used in both places. 10 tests document every config-default trap and verify parity on every bar.
Python Algorithmic Trading IBKR Backtesting pytest
- 📟 Ringora — Multi-tenant SaaS CRM & power dialer: Telnyx WebRTC, Stripe billing, AI call analysis, Rust WASM
- 🛠️ daily-builds — a public log of what I ship most days. Project repos stay private; the cadence is public.
- 🌱 openai-chart-data-analyzer — one of my first builds (2022): a PyQt GUI querying GPT-3 on chart data. Kept as a marker of how far the work has come.
- Shipping production software — SaaS platforms, telephony systems, AI pipelines, and trading bots running in prod
- AI-native development — building with Claude, Gemini, and custom multi-agent systems to move faster
- Voice & telephony engineering — WebRTC, Twilio, Telnyx, Vapi — browser-based calling is my thing
- Automating everything — if it can be automated, it should be
Most projects are closed source and in active development. If you're interested for hiring, collaboration, or licensing — reach out. Open source contributions are marked above.