I'm a backend engineer in New York who works on messaging infrastructure and observability. I studied Statistics & Machine Learning at Carnegie Mellon and AI at Duke.
- 🔧 Recent work: a zero-downtime migration of a JMS messaging platform from Tibco EMS to Solace, priority-aware message dispatching, and monitoring with Grafana / Prometheus / Loki
- ⚔️ Side project: building a practice-analytics app for Sekiro: Shadows Die Twice bosses, and teaching it to read gameplay recordings
- 🌱 Currently learning: React + TypeScript and frontend testing
- 💬 Ask me about: JMS and Solace, observability, or how to beat Genichiro
⚔️ Featured: Sekiro Boss Practice Analytics
A full-stack app that tracks Sekiro boss attempts and shows which moves keep killing you.
- Product: React + TypeScript frontend, FastAPI + SQLAlchemy + PostgreSQL backend, accounts with server-side sessions, per-user analytics, English / 中文 interface, Docker Compose and four required CI jobs
- Combat ground truth: Genichiro's moveset rebuilt from the game's battle AI, animation data and dev-menu recordings instead of wiki lists: 63 moves from 45 AI chains, with every one of the 353 animation entries reviewed
- Now: reading gameplay recordings automatically. The prototype detects when the player takes damage or has their posture broken and names the boss move behind each failure. On recordings it was never tuned on, it ranked the right move first for 36 of 41 failures; validation on more recordings is under way
| Project | What it is |
|---|---|
| Veiled Chess | A chess variant where every piece except the king starts hidden, with an engine that recommends moves under that uncertainty (Duke AIPI 540) |
| 2048 | 2048 in Python / tkinter, with an expectiminimax AI and a reinforcement-learning AI |
| Reading Comprehension with Small Transformers | Multiple-choice reading comprehension with small transformer models (CMU 11-711) |
| Virus Infection Game | An upgraded version of my CMU 15-112 term project |