7+ years building production systems across fintech, IoT, and SaaS platforms. Based in Nairobi, Kenya. Remote-native. AI-augmented engineering as a daily practice.
Primary languages, frameworks, and infrastructure tools across 7 years of production systems.
Production systems built and owned end-to-end, not demos.
Built the M-PESA Daraja payment integration end to end for a Nairobi savings cooperative: STK push flows, webhook idempotency, callback signature verification, idempotency key layer, transaction state machine lifecycle, and a nightly reconciliation pipeline that automated daily manual cross-referencing by the finance team. Operated under Central Bank of Kenya regulatory compliance.
Designed and deployed production multi-agent AI systems using LangGraph.js with MCP-compatible tool integrations, RAG pipelines with vector embeddings and semantic search, binary quality gates, human-in-the-loop approval checkpoints, and LLM cost observability dashboards. Built quality control systems specifically for validating AI-generated code before production deployment.
Architected a full-stack IoT platform processing high-frequency sensor data from 1,000+ connected industrial devices. Built React and Next.js real-time dashboards with WebSocket live data, a unified design system component library, Python and Node.js backend services, Kafka event streaming pipelines, and a full observability stack with Prometheus and Grafana built from scratch.
Led 0-to-1 delivery of a multi-tenant SaaS platform in under 5 months, architecting NestJS microservices with Kafka inter-service communication, provider-scoped PostgreSQL schema isolation, configurable admin dashboards, and webhook delivery patterns for 50+ institutional clients. Achieved 5x traffic growth with zero additional infrastructure cost.
Claude Code is my primary development environment for architectural reasoning and multi-file changes. I define technical specifications, orchestrate AI agents to implement systems, and rigorously review generated code before production. I once caught a Redis connection-pool exhaustion bug in AI-generated batch code that compiled and passed tests — but would have leaked resources under sustained concurrent load. Speed without the discipline to verify is just technical debt accumulation.
7+ years of startup and scale-up engineering, no enterprise.
Open to senior full-stack, backend, platform, and AI engineering roles globally. Remote-first. Available immediately.