Anda Usman

I’m an AI Engineer focused on LLMs, intelligent agents, and making agent systems measurable in production. I build generative assistants and LLM-powered automation pipelines that perform reliably for real users.

I designed and deployed AgentMetrics, a real-time observability platform for AI agent fleets—covering costs, failures, latency, and reliability. I also built Ember, a lightweight headless browser for AI agents that runs at 17 MB idle and avoids Docker/API key requirements.

My open-source work includes Datum Agent for autonomous dataset discovery—discovering, cleaning, validating, and tagging public African datasets for open access. I also created ConvTag, an active learning conversation labelling system with a five-stage pipeline and human-in-the-loop retraining via JSONL/RLHF export.

Earlier, I drove Condia’s growth by building AI-powered quiz generation and leading SEO infrastructure that contributed to a 500% MAU increase with zero ad spend. I’ve also led product development for Library Aid Africa, managing a data and annotation team and enabling library search across African libraries.