I am a Technical Product Manager & AI Engineer from Nairobi, Kenya.

• 🔭 I work in fintech — running product delivery for a multi-module platform and engineering the AI capabilities that ship inside it.
• 🤖 I build LLM systems for production: multi-agent orchestration, RAG pipelines, and hybrid rule-based/ML architectures — with evaluation, guardrails, and audit trails as first-class citizens.
• 🎓 MSc, Information Technology — Carnegie Mellon University (Machine Learning & Data Analytics concentration, Product Development minor) · BSc Computer Science, JKUAT.
• 🚀 Co-founder of NurtureOS — an AI-powered maternal health platform for community health workers in rural Kenya, backed by a $75,000 Enabel R&D grant.
• 📊 Presented ML research on forecasting electoral violence at the Deep Learning Indaba.
• 🌱 Community Lead at Product-Ed — co-running a Senior PM Bootcamp for product managers.
• ⚡ In my free time, I like to hike, swim and discover new things. Call me a curious-cat
• 📫 How to reach me:

As a product manager — I run program delivery across multiple modules: roadmaps and delivery plans pressure-tested against real capacity, specs whose acceptance criteria cover the failure paths, RAID logs with owners and review dates, and UAT with exit criteria set before testing starts.

As an AI engineer — I make the build call on where AI genuinely belongs and where deterministic logic wins, then engineer for reliability: hallucination prevention, groundedness checks, fallbacks, and evaluation gating every ship.

• Donna — agentic contract analyzer · multi-agent pipeline (planner → parallel retrieval → synthesis → verifier) over RAG, with groundedness scoring on every answer, policy-guarded autonomy, audit logging, and a built-in eval suite
• Production movie recommender (CMU ML systems project) · led a team of 5; Two-Tower neural network at 4.5x baseline precision, 80ms p95 — Kafka ingestion, Blue/Green Docker deploys, A/B testing, Prometheus/Grafana monitoring
• Electoral violence prediction · applied ML on conflict data — presented at the Deep Learning Indaba
• ML fundamentals from scratch · neural network built in pure NumPy; LoRA-style low-rank adapters implemented in PyTorch
• Epidemic forecasting from search trends · modeling pneumonic plague dynamics with Google Trends signals

                             

LLM & agents: Anthropic & OpenAI APIs · Llama models (via Groq) · RAG pipelines (ChromaDB) · multi-agent orchestration · LangChain · prompt engineering with versioned prompts · evaluation design · guardrails & hallucination prevention