.png&w=256&q=75)
1
1
Kenya
1 year of experience
I build AI systems that think with knowledge, not just data. I'm a Software Developer specializing in AI Agents, Knowledge Graphs, and Graph Databases. I design intelligent systems that combine semantic technologies with large language models to deliver context-aware, explainable, and scalable AI solutions. My expertise spans Neo4j, RDF, OWL, SHACL, Python, and modern AI frameworks, with a focus on transforming complex, connected data into software that can reason, retrieve, and make better decisions. Whether it's architecting enterprise knowledge graphs, building autonomous AI agents, or designing semantic APIs, I enjoy solving problems where traditional software meets intelligent automation. I'm driven by one belief: the future of AI isn't just bigger models, it's better knowledge.

WakiliIntel is a legal AI startup building the first AI copilot for Kenyan advocates. We are attacking a massive inefficiency: Kenyan lawyers spend 6-10 hours per document researching kenyalaw.org, finding templates, and drafting pleadings from scratch. There is no AI tool built for Kenya's jurisdiction; existing legal AI targets US/UK common law and ignores Kenyan statutes, the Civil Procedure Rules, and local caselaw. WakiliIntel changes that. What it does. A lawyer opens a case, uploads PDFs, and starts writing. Behind the editor, the Pydantic AI agent does the heavy lifting: it searches kenyalaw.org for relevant judgments and statutory references, indexes everything into an Oxigraph RDF knowledge graph so future runs are instant with no redundant scraping, and loads a practice-area skill from a skill folder (plaint, defence, petition, submissions, etc.) that governs drafting conventions and required structure. Then it drafts the document in real time, paragraph by paragraph, directly into the editor. The advocate edits, the agent adapts. The state is shared: every change in the editor flows back to the agent, and every agent output renders instantly via AG-UI protocol. It is a true collaborative workflow, not generate-and-pray. Why AMD. All inference runs on Fireworks AI hosted on AMD GPUs. We use token-efficient routing, a lightweight model for autocomplete and a larger one for drafting, to minimize cost. The entire stack is containerized for reproducible deployment on AMD Developer Cloud with ROCm. Market. Kenya has over 15,000 licensed advocates and millions of litigants who need affordable legal documents. We monetize via M-Pesa Daraja subscriptions, Kenya's dominant mobile money platform with 30M+ active users. Our payment stub is built and ready. Current status. Fully functional: editor, agent, knowledge graph, autocomplete, multi-document threads, M-Pesa stub. 18 tests pass. Frontend compiles clean. Ready for AMD cloud deployment and pilot launch.
13 Jul 2026