AI Engineer Resident @ Qualcomm · Spatial Computing & WebXR Developer · Full-Stack Builder
B.Sc. in Computer Science & Business Administration — University of Florida 🐊
I specialize in architecting production Generative AI & agentic platforms, on-device model acceleration, and immersive WebXR spatial computing experiences.
- 🔭 Current Role: AI Engineer Resident at Qualcomm — Advancing on-device AI acceleration, edge inference optimization, neural hardware runtimes, and generative intelligence.
- 💼 Industry Background:
- AI Engineer at SoftwareOne: Architected enterprise GenAI solutions on AWS (Bedrock, LangChain, OpenSearch, pgvector) with tenant-isolated data governance, plus autonomous multi-tool query agents with LangGraph.
- Machine Learning Intern at Verifone: Developed centralized multi-site fraud detection (TensorFlow, TCNN, Optuna, ClearML); achieved +20% anomaly detection accuracy and cut training time by 50% (#1 Software Engineering Team recognition).
- Computer Vision Intern at VKIST: Engineered COVID-safe touch-free facial identification systems (PyTorch, CNN) with 87% accuracy.
- 🎓 Education: B.Sc. Computer Science, Minor in Business Administration — University of Florida (Dean's List). Go Gators! 🐊
|
4K square spatial video extender, AI generative outpainting pipeline (LTX 2.3, Luma Ray-2, Depth Anything v2), HEVC hardware encoding, and an immersive WebXR cosmic VR player for Meta Quest 3 & Apple Vision Pro with real-time video glow (Ambilight), 6DOF spatial transport, authored Reel Packs, social feeds, and Stripe credits.
|
Autonomous enterprise query agent built with LangGraph bridging SQL and unstructured documents. Features dynamic multi-tool routing (Vector Store, SQL Engine, Web Search, Python Code Interpreter, and Data Visualization) deployed on Azure App Service with Azure PostgreSQL (pgvector).
|
|
Emotionally intelligent conversational assistant inspired by Her. Features dual processing modes: 100% private offline inference (Whisper STT + Ollama LLM + Bark TTS) or low-latency hybrid streaming via the OpenAI Realtime API (<0.5s response time) with dynamic emotional vocal modulation.
|
Multi-agent research framework using CrewAI and LangChain. Implements intelligent switching between local DeepSeek-R1 (1.5B via Ollama) and OpenAI o1-mini / GPT-4o, with web search orchestration and automatic structured synthesis reports.
|
|
Real-time emotive 3D digital avatar interface. Leverages Three.js and React Three Fiber with custom Web Audio FFT frequency analysis for real-time Oculus-style viseme mouth morph targets, procedural eye kinematics, and ElevenLabs realtime voice streaming.
|
Centralized fraud detection pipeline across 16 retail sites using Temporal Convolutional Networks (TCNN). Boosted anomaly detection accuracy by 20% and cut training duration by 50% via Optuna Bayesian optimization and ClearML monitoring.
|
|
Lightweight automation web utility for VR creators to standardize filename syntax and inject stereoscopic 180° / 360° metadata compatible with Skybox VR, Playa VR, Pigasus, and Oculus Video without re-encoding.
|
Interactive personal portfolio website featuring dark/light aesthetic modes, project changelogs, and an embedded AI assistant backed by Supabase pgvector semantic search.
|
🤖 AI, Machine Learning & Agentic Systems
🥽 Spatial Computing, 3D & Multimedia AI
⚡ Full-Stack & Cloud Architecture
-4169E1?style=flat-square&logo=postgresql&logoColor=white){kind=icon}
🐊 G O G A T O R S 🐊
_ _
/o\/o\-. _.-'^'^^'^^'^^"^^'-.
.OO.----'\ /\ / `-' /^ ^^-._
( ` \ | _ ^^-._
VvvvvvvVvv`___...)_/ /_/_/_/_/_/_/_/\ (__________^^-.
`^^^^^^^^` / / > > _`) )
(((` (((` '--'^
🧡 "In all kinds of weather, we'll all stick together!" 💙
Proud University of Florida Alum · Original ASCII Art by Joan G. Stark
Let's build something extraordinary. Reach out via LinkedIn or Email.