Viewing profile — muhammadwaqar1

AI architect in Dallas. I lead the AI engineering team at Supreme Lending, running agentic systems in mortgage production under real audit and compliance constraints, coordinated agents with persistent memory, cited outputs and human-in-the-loop gates, plus a multi-modal RAG platform serving 1,000+ branch users.

I open-source the tooling I wanted while building that:

tokenprof: profiles what is actually consuming your context window, per turn, per tool. https://github.com/muhammadwaqar12/tokenprof

awesome-agent-failures: production failure modes for LLM agents: symptoms, reproductions, mitigations. https://github.com/muhammadwaqar12/awesome-agent-failures

Mostly here for context engineering, agent evaluation and trajectory replay, and where LLMs quietly break on procedural knowledge.

Open to AI engineering and forward-deployed roles. m_waqar@live.com - github.com/muhammadwaqar12