Product leader building
production AI systems.

Fin-Ops Travel Planner: Multi-Agent Trip Budgeting Assistant

A 5-person Andela AI Engineering Bootcamp capstone squad's CrewAI system that plans a day-by-day, budget-constrained trip. My contribution: the custom MCP tools that ground the agents in real flight, exchange-rate, and destination-cost data.

Planning international travel usually means juggling separate tabs for flight prices, currency conversion, and daily budget guesswork with no single view of whether the trip is actually affordable. This was a 5-person squad capstone project: a Planner Agent, a Budget Profiler Agent, and a Flight + Forex Agent coordinate over CrewAI to produce a day-by-day itinerary and a full budget table for a given destination, dates, and total budget. My specific role on the squad was building the custom MCP tools the agents call to get real data; everyone else's individual work (agent logic, MCP integration, the Gradio UI, overall architecture) is credited to the rest of the team in the repo.

Hugging Face Spaces go to sleep after a period of inactivity. If the demo above doesn't load right away, give it a minute to wake up and reload. After it restarts, close the build log console to see the full app interface.

Litigation Prep Assistant: Multi-Step RAG Pipeline

AI-assisted litigation preparation for Kenyan law: turns case text or a PDF into a structured legal brief, streamed step-by-step to the browser in real time. A research and productivity aid, not a substitute for qualified legal counsel.

Litigation prep involves extracting relevant facts from case materials, grounding arguments in the right statutes, and drafting strategy documents: usually a slow, manual research process. This started as an Andela AI Engineering Bootcamp team capstone, where my contribution was the FastAPI backend, pipeline orchestration, and LLM integration, plus the RAG implementation I additionally took on as the project evolved (extraction → RAG-grounded strategy → drafting → automated QA, streamed to the browser over Server-Sent Events). Each teammate was then individually responsible for deploying their own version: this deployment, including the full port from the original AWS stack (App Runner + Aurora + Terraform) to a free-tier stack on Vercel, Google Cloud Run, Neon, and Pinecone, was done entirely by me to minimize hosting cost.

Digital Twin: Conversational AI About Me

An AI twin recruiters can talk to directly about my background, skills, and projects, grounded in my actual experience rather than generic chat. Built and deployed twice, on two different clouds, as a deliberate portfolio strategy.

Recruiters and hiring managers often just want a quick answer, not a full conversation. This is a conversational digital twin, grounded in my actual experience, that can answer questions about my background and projects directly. It exists as two independent, fully-built deployments of the same product on two different clouds: an AWS/Bedrock/Lambda version (not currently live, but fully documented) and this GCP/Gemini/Cloud Run + Vercel version, which is the one running now, specifically to demonstrate the same IaC and CI/CD patterns applied portably across providers, rather than tied to one cloud's specifics.

The AWS/Bedrock deployment is intentionally kept offline to avoid ongoing cloud costs. The live demo above runs on the GCP + Vercel deployment instead.