AI engineer at Microsoft with three years of experience in search and recommendation for Microsoft 365 Copilot (2M+ monthly enterprise users). I build LLM applications, hybrid retrieval and agents, and the evaluation, experimentation and telemetry that show whether they work. BSc Computer Science, First Class Honours.
I'm an AI engineer. My foundation in search, ranking and evaluation at Microsoft taught me that the hard part is rarely the model. It's knowing whether a change actually helped.
I graduated top of my class in Computer Science at the University of Nairobi, and was a 2023 Deep Learning Indaba scholar in Accra.
Failures usually live upstream of the ranker: in how documents were modelled, what was indexed, and whether the query was ever a search query.
I speak English and Swahili fluently, and some French. Now I build LLM apps, retrieval and agents end to end.
Hybrid BM25 + ANN + transformer reranking over 8,807 Netflix records behind a cached FastAPI service, with latency benchmarks and a tool-calling agent with grounding checks.
Shows: retrieval, serving, evaluation, tests, CI and Docker.
An offline CLI that triages photo folders with rules first, a local vision model second, and logs every decision. Nothing is deleted.
Shows: pipeline design, safe automation, schema-validated model output.
SQL-first analysis and chronologically validated models, explicit about prediction versus intervention.
Shows: SQL, data quality, leakage control, calibration.
Completed a 12-session mentorship supporting a university student with résumé development, networking, career planning and professional skills.
Read the reflection →
Shared lessons from working in technology with students and early-career professionals at a community event for future engineers and product leaders.
See the event post →
Part of a global community of leaders supporting women in technology through visibility, connection and community participation.
Assessed and scored technology projects submitted by Technovation teams from around the world.
Articles on what I'm building and learning in AI engineering.
An open-source AI experience for conversations with cultural legends such as La Llorona and Luanda Magere.
A safe, local-first AI assistant that organises photos without being allowed to delete anything.
More weekly writing on reliable AI applications, retrieval and evaluation is coming.
Open to roles anywhere, remote or relocation. Based in Nairobi, Kenya.
