Portfolio
AI automation, agentic systems and full-stack AI projects I've built from Lagos, spanning n8n workflows, a lead-research agent and applied ML, alongside backend and data science work at 4orge, Merrowgate and Synergy Solutions IMC LTD.
Selected Work
Most AI tools ship English-first and assume fast internet. FarmBuddy gives Nigerian smallholder farmers advice grounded in their own soil, crops and local weather, built to work in low-data areas. It's multimodal and multilingual: farmers can type, speak or send a leaf photo in Nigerian English, Hausa, Igbo or Yoruba.
Full-stack · AI, Backend & Frontend
An internal tool that turns discovery-call notes into a client-ready proposal, with Claude drafting one section at a time. A proposal can't reach the client until both the salesperson and a separate admin have approved it.
Full-stack · AI Document Generation
Inbound leads arrive as messy CSVs with no consistent format. This pipeline cleans them, removes spam and non-buyers before spending anything on AI, and gives sales a ranked list with the reasoning behind every score.
Full-stack · Data Pipeline & Applied ML
Koya AI Academy's sales, delivery and people-ops data lived in three separate systems. I built an n8n workflow that calculates every metric in code, then has Claude write the executive summary from the finished numbers.
Workflow Automation & Analytics
Koya Talent's content team produced every piece by hand, one slow step at a time. I built an event-driven system that turns an idea into source-backed drafts and channel-ready posts, with a manager signing off before anything is queued.
Workflow Automation · AI Content Pipeline
Novus Realty receives invoices by email in all sorts of layouts, sometimes attached and sometimes written into the message. I built an n8n workflow that identifies the real invoices, extracts the key details with Claude and logs every outcome to a Google Sheet, failures included.
Workflow Automation & AI Integration
Koya Talent researched outbound leads and wrote every message by hand, one company at a time. I built an agent that turns a one-line objective into an evidence-backed lead list with outreach already drafted. It can't send anything on its own.
Full-stack · AI Agent Engineering
A task manager that plans your week for you. It fits pending tasks into the coming weekdays based on deadline, priority and effort, and uses Gemini to suggest what to work on next.
Full-stack · AI, Backend & Frontend
Most heart-disease risk tools rely on lab tests that many people can't easily get. I built an explainable model that estimates coronary heart disease risk from self-reported answers alone, reaching 0.815 AUC-ROC.
Applied Machine Learning
A health-monitoring app for logging daily vitals and keeping track of medication. It flags readings outside healthy ranges and emails a caregiver when a dose is missed.
Full-stack · AI, Backend & Frontend
How I build
Most of what's here started as a personal project rather than a client brief: a problem I noticed (farmers priced out of AI tools, a health check that needed lab access) and then built end to end, from the data model and API layer through to the interface people actually touch.
Happy to talk through the architecture and trade-offs of any of them, including the parts that didn't work the first time.
Let's talk
Whether it's an AI system, a backend platform or a research collaboration, let's start a conversation.