Hire Me
Computer Science graduate and software developer focused on AI deployment and web development. US citizen, eligible for government security clearance.
Summary
Computer Science graduate and software developer focused on AI deployment and web development. Designs, builds, and self-hosts production websites end to end — front end, backend, Linux ops, and on-prem LLM integration via Ollama. Production experience supporting AI-integrated enterprise software, with side projects spanning real-time inference benchmarking, native iOS apps, and live web tools. Comfortable across Python, JavaScript, Swift, and the systems stack underneath them.
Skills
AI / ML
PyTorch · ONNX Runtime · NVIDIA TensorRT · CUDA · computer vision (YOLOv8) · FP16 / Tensor Cores · ONNX export · TRT engine build · latency benchmarking (p50/p95/p99) · local LLM integration (Llama 3 / Ollama)
Languages
Python · JavaScript (ES6+) · Java (OOP) · Swift (SwiftUI) · HTML5 · CSS3
Systems & Tooling
Linux (Fedora) · Apache · Git / GitHub · VS Code · PyCharm · Xcode · Cursor · self-hosted deployment · reverse proxying · DNS
Engineering practices
Object-oriented programming · data structures & algorithms · concurrency · I/O · debugging · root-cause analysis · structured documentation · Agile / Kanban (Jira, Trello)
Selected Projects
trtbench — Multi-Runtime Deep Learning Inference Benchmark
- Built an end-to-end deployment pipeline for a YOLOv8n object detector and benchmarked inference latency and throughput across six runtime / precision configurations on an RTX 4080.
- Achieved 1,211 FPS at 1.15 ms p99 latency with TensorRT FP16 — a 3.0× throughput improvement over PyTorch eager-mode CUDA on identical inputs.
- Authored the ONNX export pipeline and TensorRT engine builders (FP32 and FP16) from PyTorch source weights; verified ONNX numerical parity within ~1e-4 of source and confirmed TensorRT functional parity on real images.
- Designed a rigorous benchmark protocol — 20-iteration warmup, explicit CUDA stream synchronization, 200 timed iterations per configuration, p50 / p95 / p99 reporting — isolating pure forward-pass cost from preprocessing and NMS overhead.
- Captured reproducible system metadata (PyTorch / CUDA / TensorRT / ORT / GPU / OS) per run and produced cross-platform comparison tables and latency-distribution charts; scoped future work including INT8 calibration and Jetson edge benchmarks.
frontendneeded.com — Self-Hosted Site with Local LLM Integration
- Designed, built, and operate a 24/7 self-hosted Fedora Linux + Apache deployment; manage DNS, deployment workflow (Git-pull on server), and uptime.
- Integrated a local Llama 3 LLM via Ollama, reverse-proxied through
Apache, exposing an
/api/chatendpoint that powers the on-device AI chat on this page — fully on-prem inference, no third-party API dependency.
Tarella Notes
Designed and built a lightweight iOS notes app from scratch. Handled the full lifecycle — UI design, data modeling, and App Store distribution. Iterated on features based on real user feedback.
Warframe Price Checker
Built an asynchronous Python tool that fetches bulk item pricing data from the Warframe.market API, parses JSON responses, and surfaces the highest buy price per item for the user.
Experience
Technical SEO Specialist / Web Developer
- Conducted technical audits on a large-scale healthcare platform, debugging crawl, indexation, and structured-data issues at production scale.
- Designed and shipped a consumer-facing comparison interface; built and maintained scalable state-specific landing pages with CTA components driving internal linking and conversion.
Software Support Technician
- Subject matter expert and escalation point for two AI-integrated enterprise products (ScheduleAI, NoteGuardAI) within a healthcare-domain platform.
- Applied structured debugging methodologies to diagnose application, configuration, and system issues; partnered with engineering and product to escalate and document root causes.
Education
B.S. Computer Science
Ask about me
Hey! This is an AI assistant that knows about Key's background, skills, and projects. Ask me anything you'd like to know.