I'm a computer science student at NYIT (graduating 2026), currently working as a Laboratory Consultant at the NYIT ETIC Lab. I work across the stack but I'm most drawn to AI/ML in production: models that retrain themselves, get monitored, and actually ship.
- I run a production ML pipeline that predicts NBA/MLB game winners and player props with an XGBoost + LightGBM ensemble: test-set accuracy, 50+ engineered features, and a fully automated Celery/Optuna retraining loop with MLflow-tracked champion/challenger promotion, running on a self-hosted GPU server. The feature engineering and model training code is open-sourced at sports-prediction-model.
- I built rmbr, an embedded, local-first memory + retrieval engine for AI agents: one SQLite file, no server, no API key required. Published on PyPI, listed as an MCP server on the official MCP registry and on Glama.ai, with an optional HTTP mode too.
- I built pyhomerun, a zero-dependency Python library for baseball stats: sabermetrics and an MLB Stats API client, published on PyPI.
- On the side, I develop custom Lua & C# scripts for games like Holdfast, GTA, and FiveM, which is where I learned to read messy systems, reverse-engineer APIs, and ship for real players.
- My current hobby project is SoundKitten, a lightweight Rust/Tauri desktop SoundCloud client for people who'd rather not run a browser wrapped in an "app." Source at soundkitten.
|
An embedded, local-first memory + retrieval engine for AI agents: one SQLite file, no server, no API key.
|
A lightweight Rust/Tauri desktop SoundCloud client for people who'd rather not run a browser wrapped in an "app."
|
|
A clean, dependency-free Python library for baseball stats: sabermetrics + an MLB Stats API client.
|
Open-source feature engineering + XGBoost/LightGBM training code powering my production NBA/MLB winner & player-prop models. Self-retrains nightly via Celery + Optuna, MLflow-tracked.
|
|
The top tier of the Cipher line, tuned for quality over speed on higher-end hardware.
|
The mid tier, balancing speed and accuracy for local assistants on consumer hardware.
|
|
The lightest tier, sized to run fully on-device for fast, low-resource triage.
|
|
I'm an Early 🐤
🌞 Morning 1658 commits ███░░░░░░░░░░░░░░░░░░░░░░ 13.07 %
🌆 Daytime 5568 commits ███████████░░░░░░░░░░░░░░ 43.90 %
🌃 Evening 4600 commits █████████░░░░░░░░░░░░░░░░ 36.27 %
🌙 Night 857 commits ██░░░░░░░░░░░░░░░░░░░░░░░ 06.76 %
📅 I'm Most Productive on Saturday
Monday 2027 commits ████░░░░░░░░░░░░░░░░░░░░░ 15.98 %
Tuesday 1502 commits ███░░░░░░░░░░░░░░░░░░░░░░ 11.84 %
Wednesday 1578 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.44 %
Thursday 1778 commits ████░░░░░░░░░░░░░░░░░░░░░ 14.02 %
Friday 1491 commits ███░░░░░░░░░░░░░░░░░░░░░░ 11.76 %
Saturday 2285 commits █████░░░░░░░░░░░░░░░░░░░░ 18.02 %
Sunday 2022 commits ████░░░░░░░░░░░░░░░░░░░░░ 15.94 %
Last Updated on 02/10/2026 11:48:04 UTC
Open to internships & new-grad roles