CS student at FAST-NUCES Islamabad (graduating '27) who builds AI systems that actually have to work — not just pass a demo.
Currently interning at XCorre architecting LLM pipelines for live brand campaigns. Before that I learned the hard way that a prompt that looks great in a notebook can quietly destroy your output quality in production. Now I maintain versioned prompt libraries and fallback behaviors like a religion.
🔭 Currently working on
- Agentic AI pipelines for automated marketing at XCorre (LLMs + GANs in production)
- AI automation systems: voice agents, lead gen workflows, social media optimization
🌱 Currently learning
- Production-grade LLM evaluation and regression testing
- CI/CD and cloud fundamentals for AI systems
- Making multi-agent systems less chaotic and more trustworthy
🏆 Recent win
- 🥉 Ranked 3rd nationally in the All Pakistan Prompt Engineering Competition out of 3,000+ — took home PKR 50,000 and a lot of respect for how hard Pakistan's CS community goes
💬 Ask me about
- Agentic workflows and where they break in the real world
- Prompt engineering that actually holds up under pressure
- Building with Claude API / MCP / n8n
- Getting AI systems from "it works on my laptop" to something other people can use
🤝 Looking to collaborate on
- Open-source AI tooling (already contributed to a 26,500 ★ repo — it's addictive)
- Anything at the intersection of LLMs and real-world automation
📫 Reach me at
⚡ Fun fact
- I co-founded an education non-profit at 14. We've run robotics and AI programs across 5 institutions in Pakistan. Still going.
- I also deadlift. Started at 80kg two years ago. Now at 180kg. Consistency works the same way in the gym and in the terminal.