Problem

LLM-powered products ship changes to prompts and models with no systematic record of what changed, when, or why. Regressions surface in production. Bias goes undetected until a user reports it.

Approach

Plugged into GitHub Actions as a CI gate. Every prompt change triggers automated testing, drift detection, and a compliance certificate before merge. The pipeline rejects non-compliant changes before they reach production.

Problem

Cross-border African trade runs on trust that infrastructure cannot yet verify. Buyers and sellers transact blind — no escrow, no identity layer, no dispute resolution.

Approach

Built identity verification, escrow payments, and real-time chat into the core transaction flow — not as add-ons but as the product. No trade completes without the trust layer.

Problem

Nigerian pharmacy chains lose significant margin to expired stock that was not flagged in time. Inventory tracked in notebooks or WhatsApp threads cannot alert before the damage is done.

Approach

Built expiry tracking as the core feature — not a report, a live alert. Every SKU has an expiry date. The system surfaces at-risk stock 30, 14, and 7 days out with branch-level visibility.

Problem

A church, a clinic, and a restaurant share operational patterns — scheduling, member records, billing, reporting — but no single platform covers all three without forcing each into a wrong-shaped box.

Approach

Built a modular multi-tenant core that re-skins into industry editions: ChurchOS, ClinicOS, RestaurantOS. The schema and permissions layer is shared. The UI and terminology surfaces are configurable per vertical.

Problem

Nigerian SME owners manage finances across WhatsApp threads, paper receipts, and disconnected bank apps. Month-end is archaeology. Tax season is guesswork. Payroll takes a Sunday.

Approach

Built payroll, invoicing via WhatsApp, bank sync across 20+ Nigerian banks, and tax return generation into a single platform. Payroll calculates PAYE to FIRS 2024 spec. Invoices carry a Paystack payment link by default.

Problem

Enterprise vendor onboarding for ESG and trade compliance runs on email attachments, manual checklists, and PDF uploads that no system can verify automatically. Compliance officers chase documents. Vendors repeat the same submissions for every buyer.

Approach

Built a multi-tenant onboarding system where vendors upload once and buyers verify cryptographically. AI cross-references invoices, bills of lading, and ESG documents against compliance requirements without human review of every line.