"Do or do not. There is no try." — but I do try, then debug, then try again.
Full Stack Dev · Distributed Systems · MS CS @ UH
- Building tools that feel like magic and systems that don't lie about consistency
- Heavy into distributed systems and AI/ML — this is where I live right now
- Always thinking about the next side project
AI & Agents
-
Agentic Scrum — Multi-agent bug-fixing pipeline where PM, Dev, and QA agents (played by The Office characters) collaborate to patch your code. Built with Gemini 2.5 Flash, plain Python, zero frameworks.
-
LLM X-Ray — Visual debugger for large language models. A 7-panel interactive dashboard that makes tokenization, embeddings (3D star map), attention, temperature effects, hallucination detection, and context windows all visible and interactive in real time. React + FastAPI + Gemini API.
-
Browser Agent — Automates real browser tasks with plain English. Solves captchas, takes screenshots, runs your errands.
Productivity & Tools
-
Silent Meeting Transcriber — Transcribes Teams meetings (and any audio call) without notifying participants. No bots, no cloud uploads. Runs entirely locally with OpenAI Whisper. Transcripts go straight to a
.txtfile you can paste into any AI. -
Notably AI — AI-powered note-taking workspace with real-time collaboration, comments, and a polished UI.
Distributed Systems
-
CAP Store — Hands-on CAP theorem demo: a two-node key-value store you can flip between CP and AP mode to watch the trade-offs play out in real TCP traffic. Last-write-wins today, vector clocks next. Zero dependencies beyond stdlib.
-
Hardware CI Simulator — Simulates a hardware-in-the-loop CI pipeline for HPC kernels: node provisioning, firmware flashing, NCCL/RDMA testing, regression detection, and parallel execution. No real hardware needed.
Just for Fun
- Pixie Chat — PHP + MySQL real-time local chat app. A nostalgic experiment that got out of hand.
The stuff that's actually moving right now:
- Inference-time compute and reasoning models — how thinking longer beats scaling bigger
- Multi-agent orchestration and where reliability constraints dominate system design
- MoE architectures and how sparse activation changes what's possible at scale
- Distributed consensus — Raft, vector clocks, CRDTs, and what production systems actually do
- Long-context and memory-augmented systems
"When 900 years old you reach, ship as many side projects you will not." — Yoda probably

