// AI Engineer · IBM · Islamabad, Pakistan
Syed Asad
Zaman
I build LLM systems, RAG pipelines, and autonomous agents for enterprise environments. Three production systems. Real workloads. Boring problems solved well.
Currently: deploying GenAI tooling on top of enterprise asset data at IBM. Writing about what actually works in production AI.
// latest posts all posts →
- Single-Agent Is Not the Answer. It's the Current Shortcut. Why defaulting to single-agent architectures makes sense today, but treating it as a permanent principle is shortsighted.
- The AI Layoff Problem Is Not What You Think Why the AI layoff wave is a coordination failure that hurts firms too, and why UBI, capital taxes, and retraining can't fix it.
- Stop Building More AI Agents. Start Packaging Expertise. Most AI agents are the same runtime in different clothing. The real unlock isn't a better shell — it's packaging reusable expertise that agents can actually use.
- Picking the Right ML Model Isn't Magic. It's a Process. A practical guide to model selection in machine learning—why train/validation/test splits matter and how to avoid picking a 'lucky' model.
- How Machine Learning Projects Actually Work: A Practical Guide to CRISP-DM A clear, practical breakdown of the CRISP-DM framework for machine learning projects, from business understanding to deployment.