I build full stack web products end to end — typed React/TypeScript frontends, Node.js backend services, and the cloud infrastructure they run on. For the last several years I've been the full-stack & cloud engineer behind an IoT platform serving 55,000+ active devices on Google Cloud, where reliability, scale, and clean architecture aren't optional.

I like choosing the right tool over reinventing the wheel, designing before I build, and shipping things that stay up.

💼 Open to the right senior full-stack / cloud role — fastest way to reach me is email. I also take on contract & consulting work through my software studio, OwnStack.

📫  nilushan.silva@gmail.com  ·  LinkedIn  ·  nilushansilva.info

Senior Cloud Engineer @ Zimi Ltd — sole full-stack & cloud engineer for a smart-home IoT cloud product (Dec 2020 – Present)

• 🚀 Migrated a live IoT platform of 55,000+ devices from Xively to GCP with zero downtime — 50% lower infra cost, 40% lower latency, 99.999% uptime
• 🏗️ Designed a service-oriented architecture on GCP (Kubernetes · Cloud Functions · PubSub · Cloud SQL · Redis) handling 100+ events/sec
• ⚛️ Built React/Redux admin dashboards and high-performance Node.js/TypeScript REST + event-driven APIs
• 🔌 Engineered and certified Google Assistant & Alexa voice integrations; designed authenticated 3rd-party APIs
• 🔧 Set up CI/CD with Docker · Cloud Build · GitHub Actions · Kubernetes across multiple environments

I built my craft the hard way, before AI tooling existed: researching from first principles, reading documentation cover to cover, and personally designing, implementing, and testing every part — and automating its delivery with CI/CD. That foundation means I understand why a system behaves the way it does — not just how to wire it together.

I bring that same rigour to agentic development — engineering, not vibe coding. I plan the architecture and choose the stack up front, then drive AI tools with spec-driven and test-driven development: clear specifications and tests come before implementation. I give agents precise, well-scoped instructions — naming the tools, patterns, and architecture to use — and lean on reusable agent skills, subagents, and deliberate context engineering to keep output consistent, reviewable, and production-grade. The hard-won engineering judgement keeps the work correct and trustworthy; the tooling keeps it fast — depth and velocity.

I'm also building with agents, not just coding with them — exploring agentic systems and multi-agent orchestration with LangGraph, function-calling / tool-use models (e.g. Nous Hermes), and self-hosted LLM deployments to design, run, and evaluate autonomous workflows.