About


I’m Salman Quazi, Director of Engineering for Microsoft CoreAI in Mountain View, CA. I’ve been building software professionally for over 20 years, across distributed systems, developer tooling, and — more recently — large language model inferencing & agents.

This blog is where I work through ideas that interest me: how LLMs actually work under the hood, software design patterns worth knowing, and whatever else I’m figuring out.

Current Work

At Microsoft CoreAI, I lead engineering teams working on Foundry Agents and developer experiences. Before Microsoft, I spent years building enterprise software across a range of stacks — .NET, distributed systems, and cloud infrastructure.

Areas of Focus

  • LLM internals — tokenization, attention, constrained decoding, function calling, agentic patterns
  • Software architecture — design patterns, complexity, benchmarking, systems optimization
  • Engineering craft — what separates good engineering from the rest

Background

  • 20+ years of professional software engineering
  • California State University, Northridge

Elsewhere