Built end to end.

Penlabs Lighting Calculator

Lead Engineer · Pen Homes

A production lighting-design suite that turns room geometry into IES-compliant plans, costed bills of materials, and qualified leads.

The problem

Lighting a home is usually guesswork: fixtures chosen by taste, not by whether a room actually hits the light level it needs. The tooling to do it properly lived in expensive desktop software, out of reach for the way homes get specified in Ghana.

What I built

• A lumen-method calculation engine with a delivered-illuminance check that back-calculates actual lux against IES targets per room type, so plans are provably correct, not just plausible.
• Layered lighting (ambient / task / accent) mapped to real fixture families, plus an interactive canvas Room Designer for arbitrary floor-plan polygons with automatic fixture placement and beam-coverage visualisation.
• A currency- and market-aware cost/energy/ROI engine (Ghana GHS + international USD) with a passcode-protected, HMAC-secured admin for live price and tariff overrides.
• A branded PDF report generator, an end-to-end lead-capture funnel with transactional email and rate-limited APIs, and full offline support as a PWA.

• Next.js 15
• TypeScript
• Prisma 7
• PostgreSQL
• @react-pdf/renderer
• PWA

PenVault

Lead Engineer · Pen Homes

A production, Cloudflare-native AI knowledge platform: retrieval-augmented generation with grounded content agents.

The problem

A growing body of internal knowledge is worthless if it can't be found and reused. PenVault turns raw documents into a searchable, grounded knowledge base that can draft trustworthy material on demand.

What I built

• A Cloudflare-native RAG pipeline: R2 event-driven ingestion, queue-based processing, markdown chunking, Workers AI embeddings, and metadata-filtered vector search on Vectorize.
• An agent engine that drafts grounded content (blog, social, scripts) from the knowledge base via LLM APIs, with content-hash deduplication and incremental indexing.

• Cloudflare Workers
• R2
• Queues
• Workers AI
• Vectorize
• RAG
• LLM APIs

CVVoice

Personal

A WhatsApp-based CV and cover-letter generator built for mobile-first users. No app to install, just a chat.

The problem

Most CV tools assume a laptop and a stable connection. For a lot of people the phone is the only computer, and WhatsApp is the internet. CVVoice meets them there.

What I built

• A WhatsApp conversational flow that generates CVs and cover letters, including voice notes transcribed with Whisper.
• PDF and Word document generation, backed by queued processing and integrated payments (Paystack, Hubtel).

• PostgreSQL
• Prisma
• Redis
• Bull
• Whisper
• Paystack
• Hubtel