A script makes this card each day. See scripts/activity_card.py. It has zero dependencies. It reads the GitHub API directly.
I am an MSc student in Computer Science at Ontario Tech, in the UIXID/CodeandSocery Lab. My areas are software design, AI, and reinforcement learning. I hold a BIT in Networking and Cybersecurity, with Dean's List, and the CCNA. I build deep RL agents. I also program Intel Tofino P4 data planes, and do Network Automation.
RIDGE — State-Conditioned Reward Blending for Behavioral Coverage in Deep RL Game Agents · IEEE CoG 2026, Madrid K. Chua, A. M. M. Shifan, A. Neshati, L. Zaman, C. Politowski
One PPO agent plays four personas. Sigmoid functions blend the personas in real time. The input is a six-value game-state vector.
| RIDGE | PPO agent with four blended personas, on Crafter | PyTorch PPO |
| PEAK | DRL benchmark engine · more than 1000 env steps/s | Gymnasium SB3 |
| Local-AI-Dungeon | offline text adventure on a local LLM | Ollama Gemma 3 |
| steam-patch-notes | clean Steam patch-note dataset and pipeline | Python API |
| Procedural-Graph-Tools | graph-based procedural generation | GDScript |
| Gestura | gestures become SOAP clinical notes · HackHive 2026 | Flask Three.js |
| Re;Animate '26 | 🥇 tactics game in AMOS BASIC, on an Amiga | chaos |
I made the lab's first Intel Tofino P4 deployment. It reads per-flow telemetry at line rate. I repaired the P4 control-plane servers. Crashes decreased by approximately 30 percent. I found 15 exploitable faults in a live WooCommerce + Stripe stack. I was a TA for approximately 120 undergraduates in two networking courses.
Open the full list
Languages
AI / ML
Networks / Security
Systems / Tooling
Learning
The best software feels like a game.
