CS @ Texas A&M · Brown Engineering Honors · 4.0 GPA
Building AI products, developer tooling, and quantitative systems.
I'm a Computer Science student at Texas A&M University (Class of 2029) in the Craig and Galen Brown Engineering Honors program. I ship end-to-end systems — from LLM-powered SaaS and coding-agent infrastructure to low-latency market bots and computer-vision pipelines.
Right now I'm focused on agent coordination, AI product engineering, and quantitative modeling.
| Education | B.S. Computer Science, Texas A&M · Brown Engineering Honors · 4.0 GPA |
| Roles | Full-Stack Developer @ Finch / ApplyEasy · Quantitative Developer @ Maroon Fund ($70K+ AUM) |
| Interests | Coding agents, realtime systems, ML/CV, prediction markets, mobile |
|
Realtime coordination for coding-agent teams Published
|
Claude · Codex · Cursor limits in the Mac notch Native Swift macOS app — live usage gauges and sessions for Claude Code, Codex, and Cursor, rendered in the notch. Local-only, no Electron, no cloud.
|
|
Issue → fix → PR agent loop Autonomous watcher that polls GitHub issues and dispatches Cursor agents to patch code and open PRs — with humans still in the loop.
|
Two-stage acne CV pipeline YOLOv8m lesion detector → EfficientNetB0 type classifier → rules-based routine mapping. Learning project (not medical software).
|
|
AI job-application SaaS Five-repo platform: React on Cloudflare, Flask / LangChain backend, Manifest V3 extension, Python CLI — deployed on AWS EC2 with nginx, SSL, and GitHub Actions.
|
Quant research for a student fund XGBoost buy-the-dip / sell-at-peak classifiers with stationary features, TimeSeriesSplit, and walk-forward validation for a $70K+ AUM fund.
|
More projects
| Project | Stack | Highlights |
|---|---|---|
| Southwest Spoilage Model | Python, XGBoost, notebooks | Time-honest 3-class model for airline crew-sequence spoilage — calibrated P(spoil) for schedulers |
| High-Freq Arbitrage Bot | C++20, WebSockets | Polymarket / Kalshi bot with lock-free feeds, TLS pooling, and ns latency histograms |
| TAMUSkate | React Native, Supabase | Campus skater social app — maps, heat zones, Google Auth, AI spot safety checks |
| PiPartner | iOS, OpenAI, OCR | AI math tutor — 400+ downloads, 5★, classroom adoption |
| Xcopilot | GCP, Gemini Flash | Fine-tuned personality-aware tweet generation from user history |
| Akash Chat | React Native, Vision AI | Multimodal attachments routed through vision models into the LLM |
| Carbon Footprint Initiative | Streamlit, ML | Hackathon emissions calculator with visualizations + offset actions |
Contributions merged upstream into high-traffic agent / infra repos:
| Repo | What I shipped |
|---|---|
| headroom | Token-savings KPIs for CI gating · normalized Anthropic model IDs across proxy paths |
| ECC | Closed a credential-leak vector (tokens off curl argv → stdin) · CI regression guard |
| CloakBrowser | Idle Chrome process cleanup · hardened post-scroll click targeting |
| hermes-agent | Windows→WSL cwd translation · session-origin persistence across gateway restarts |
Languages
Frameworks & AI
Infra & Tools
- Published @kumario/synapse — realtime contract-level coordination for multi-agent coding teams
- Shipped agentNotch — native macOS notch UI for Claude / Codex / Cursor usage & sessions
- Built Finch's AI application pipeline processing 100+ jobs/day with dual-LLM generation and production AWS / Cloudflare infra
- Quant models for a $70K+ AUM student fund — leakage fixes lifting sell-side ROC AUC from 0.64 → 0.97
- Upstream PRs into major agent / infra OSS (headroom, ECC, CloakBrowser, hermes-agent)
- PiPartner — 400+ downloads, 5-star rating, classroom adoption
Open to software engineering, AI / ML, mobile, and quant internship opportunities.