Full-stack engineer · IoT · Birmingham, UK
I treat domain understanding as the part of the work that cannot be delegated. AI tools handle syntax and boilerplate; I guard the parts that decide what a system is actually for — what to keep, what to discard, where the failure modes are, and how the data flow reflects the real-world problem.
Currently building deeper ML expertise on top of a full-stack + IoT foundation — UK Spouse Visa, no sponsorship required.
Lidless Platform — Multilingual, IoT-aware conversational system
React Native · TypeScript · Apollo GraphQL · Python · Anthropic Claude API · PostgreSQL
Distributed-by-design across four components (Controller / Hermes / Oracle / message broker), so a fault in one does not bring down the others. CQRS with outbox-inbox at the boundaries; GraphQL BFF gateway absorbing client divergence; tool schemas kept language-agnostic so the LLM serves whichever language the user is most comfortable in. Built with social housing tenants in mind, where first-language access matters.
OpsCheck — Home IoT telemetry & data analysis system
TypeScript · Fastify · MQTT (QoS-1) · Zigbee · RabbitMQ · PostgreSQL · AWS SNS · Python
Zigbee chosen over Wi-Fi-only sensing because the data pipeline depends on the sensor layer not being the weakest link. Outbox + idempotent inbox + QoS-1 in combination — designed for a system where data preservation matters more than throughput. Notifications fire on state transitions, not raw thresholds: humidity must hold above 90% for an hour before the system shifts state and emits an alert. Behaviour modelled on how humidity actually behaves in the room.
Owlverload — Inventory & expiry tracking, in daily use at a Birmingham supermarket
React Native · TypeScript · Go · PostgreSQL · Cloudflare R2
Sole-developer SDLC from Figma wireframes through App Store and Google Play deployment. Shipped v0 without wireframes to learn from real use; wireframed v1 after four months of observed friction. Wi-Fi dead zones in the shop were causing duplicate writes on reconnect, so an idempotency-key contract was paired with retry behaviour — interface and system contract designed together because the user's experience of failure depended on both.
| Layer | Tools |
|---|---|
| Languages | TypeScript · Go · Python |
| Frontend | React · React Native |
| Backend | Node.js · Fastify · Apollo GraphQL |
| Data | PostgreSQL · RabbitMQ · MQTT · Zigbee |
| ML / Data | Pandas · Scikit-learn · Anthropic Claude API |
| Infra | Docker · Nginx · Cloudflare R2 · AWS (foundational) |
- Domain first. Architecture serves the problem, not the other way around.
- Failure modes are part of the design, not an afterthought. Outbox patterns, idempotency keys, QoS settings — these are decisions about what the system is allowed to lose.
- MVP first, wireframe later, when the problem is messy. Real use produces friction that upfront design rarely captures.
- Honest postmortems over polished retrospectives. I write down the wrong hypotheses, not just the right diagnosis.
📍 Birmingham, UK · 🗣 English / 한국어