Currently building AI agents that actually work — not just prototypes.
I'm a CS student from Pakistan working at the intersection of Agentic AI and real-world automation. My focus has shifted from building one-off demos to shipping production-ready systems: platforms with real auth, real databases, real observability, and real users in mind.
To get there consistently, I build using loop engineering — tight, repeatable cycles of spec → build → validate → refine instead of one-shot coding. Every project starts with a clear spec (SDD-RI: Spec-Driven Development with Rapid Iteration), gets implemented in small verifiable loops, and is validated against real requirements before the next loop begins. This keeps AI-assisted development predictable instead of chaotic — I use Claude Code, custom agents, and MCP-powered tool ecosystems to run these loops end-to-end, from requirement to tested, deployable code.
- 🤖 Design autonomous AI agents that reason, plan, and act across tools and APIs
- 🔁 Ship via loop engineering — iterative spec → build → validate → refine cycles, not one-shot code
- 🧠 Apply SDD-RI: structured requirements before implementation, for predictable outcomes
- 🗄️ Build full production stacks: auth, relational + vector databases, caching, observability
- 🌱 Engineer MCP-powered ecosystems for secure, standards-based agent ↔ tool communication
- 💬 Always open to connecting with builders and researchers exploring autonomous systems
🔭 What I'm currently working on
- Building Humanoid AI Studio — a production learning platform for Physical AI & Humanoid Robotics, with a RAG chatbot, Postgres + Qdrant data layer, and full auth
- Sharpening agent-assisted debugging workflows with Live Code Base Debugging Oracle (vector search + Claude over a live repo)
- Refining my loop-engineering workflow: tighter spec → build → validate cycles for shipping production code faster
- Exploring deeper MCP (Model Context Protocol) server design patterns
⚡ Fun facts
- I plan my code before I write it — spec first, implementation second, every loop
- I'd rather debug a live agent trace than read documentation twice
- Ask me about MCP servers, I will not stop talking
Production-grade learning platform for Physical AI & Humanoid Robotics — RAG chatbot, Postgres + Qdrant data layer, Redis cache, and full auth. | Live, agent-assisted debugging over a real codebase using vector search and Claude via MCP. |
Console-based OOP application: flight booking, payments, user management, and a loyalty program. | Full-stack record-management system for pharmacy inventory, sales, and analytics. |
📁 More projects
- ⌨️ Typing Master — Educational typing game with progressive levels and performance tracking (
C++) - 🗣️ Effective Communication App — App focused on improving effective communication (
PowerShell)
💬 Let's connect and build something great together!