Riggs London Kenya
Pre-Launch · Active Build
Next.js 14FastifyPostgreSQL + pgvectorClaude 3.5 HaikuM-Pesa Daraja APITurborepo
Overview
Riggs London Kenya is a premium ecommerce platform built for the Kenyan fragrance market. Mobile-first and M-Pesa-first, it features a Claude 3.5 Haiku-powered AI Scent Advisor, semantic vector search via pgvector and OpenAI embeddings, WhatsApp Business API order notifications, and a Railway + Vercel deployment stack targeting ~KES 14,000/mo infrastructure cost.
Challenges
- Integrating M-Pesa Daraja STK Push reliably within a Next.js + Fastify architecture with webhook confirmation
- Building a semantic fragrance recommendation engine using pgvector embeddings on a cold-start catalogue
- Keeping infrastructure costs under KES 14,000/mo while maintaining production-grade reliability
Solutions
- Built a dedicated Fastify payment service with idempotent transaction handling and M-Pesa callback verification
- Generated OpenAI embeddings for fragrance descriptions at catalogue ingest time, enabling sub-100ms semantic search via pgvector
- Chose Railway for the Fastify API and Vercel for the Next.js frontend — right-sized for MVP traffic with zero idle cost
Languages & Tools
Next.js 14
Fastify
PostgreSQL + pgvector
Claude 3.5 Haiku
M-Pesa Daraja API
Turborepo
Key Stats
Week 8
MVP target
M-Pesa
Primary payment method
Claude 3.5
AI Scent Advisor model
~KES 14k
Monthly infra cost