🔭 AI Engineer · Agent Tooling · Data Pipelines · Vector AI Scholar @YorkU
🌱 Building human-approval tooling for AI agents and LLM integrations with human review 📊📈
🎓 MSc Computer Science (Artificial Intelligence), York University, 2026
💬 Ask me about Agents, LLMs & NLP.
📫 Reach out at: iampatelom@gmail.com
⚡ Fun Fact: I'm an avid podcast listener, always learning something new on the go!
👉 Check out my Portfolio and Resume for the full scoop 🍨🍨
| Demo | What you can do |
|---|---|
| Step through a real approval session between an operator and a coding agent. | |
| 🖥️ Remote Codex Control web console | Use the real console on a simulated Mac: open threads, send a message, approve or deny an action. Made-up data, no network calls. |
| 📊 TTC Pulse delay explorer | Explore 1.1 million Toronto transit delay events by hour, cause and route. |
| Project | What it is | My part |
|---|---|---|
| 🤖 Remote Codex Control | Human-approval layer for autonomous coding agents. Approve, steer or deny each action from a private Telegram chat while the code stays on your machine. TypeScript, SQLite, MIT. | Solo |
| 🧭 ClawCompass | Capability broker for agents: ranks skills, MCP servers and sub-agents, redacts secrets from context, gates paid and risky actions. OpenClaw Hack Toronto, 2026. | Team of four; I led the build |
| 🏆 Synapse AI | Multi-agent study assistant with a classroom API for teachers. FastAPI, Next.js, MCP. AgentShyft Hackathon winner, 2026. | Team of three; I built the teacher and classroom API |
| 🚇 TTC Pulse | Medallion pipeline (bronze, silver, gold) and star schema over 1.1 million TTC delay events, joined to GTFS and live alerts. Python, DuckDB, Parquet, Streamlit. | Team project; I wrote most of the pipeline |
| 🚦 Toronto Traffic Anomaly Detection | Poisson GLM versus GRU on hourly City of Toronto traffic counts, with KL-divergence anomaly scores. PyTorch. | Solo |
| 🏭 LLMs for Job-Shop Scheduling | Fine-tuning Llama 3.1 8B with 4-bit QLoRA to write schedules from text, and why benchmark-size prompts overflow the context window. | Team of three |
