Open to Principal / Head of Platform roles
Principal Programme Manager — GenAI & Cloud-Native Platforms
Strategic engineering program leader and architectural authority directing complex, end-to-end technology and infrastructure delivery across enterprise domains. Deep Kubernetes expertise spanning cluster architecture, Helm-based microservice delivery, service mesh, and production deployment pipelines — with hands-on background in backend engineering, DevOps, SRE, and GenAI/RAG systems built on Python and Java.
Proven track record bridging engineering, operations, and external partners to establish governance frameworks, manage master schedules, and de-risk critical-path deployments. Experienced in building delivery standards from scratch, standing up PMO frameworks, and managing multivendor/partner ecosystems for high-availability systems — leading multidisciplinary teams across UX, frontend, backend, QA, and platform engineering to deliver production-grade systems at scale. Passionate about architectural excellence, mentoring engineers, and enabling teams to adopt modern AI, cloud, and platform engineering practices.
I still write code. Most of my recent hands-on work sits at the join between GenAI and platform engineering — RAG pipelines and their evaluation harnesses, MCP servers that give models safe access to operational systems, multi-agent flows in Google ADK, and the Kubernetes underneath all of it. I write up what I learn in Writing, and there are three worked examples in Selected Work.
British Telecommunications (BT)
British Telecom
Vodafone New Zealand via Infosys · New Zealand
Tata Consultancy Services (TCS) · Kochi, Kerala
A closer look at three systems I've been hands-on with. The common thread: the model is the easy part. Retrieval quality, tool boundaries, and the platform underneath are where the real engineering lives.
RAG · Retrieval-Augmented Generation
Engineers were losing the first ten minutes of every incident hunting through Confluence, PDFs, and tribal knowledge. We put a retrieval layer over that corpus so they could ask a question in plain English and get an answer with the source attached.
MCP · Model Context Protocol
Before MCP we had the same integration written three times — once per assistant, each with its own auth handling and its own subtly different idea of what "get circuit status" meant. Moving those integrations behind MCP servers made them a platform capability instead of per-app glue code.
Agents · Google ADK
One prompt holding a dozen tools is a system that guesses. Splitting the work into a small set of focused agents with a coordinator on top made behaviour far easier to reason about — and much easier to test, because each agent has a job you can write assertions against.
Notes from actually shipping this stuff — mostly the parts that surprised me. I write these partly for the engineers I mentor through The AI People UK, and partly because writing it down is how I find out whether I understood it.
The demo was great. The pilot was not. Almost every problem turned out to be a retrieval problem wearing a model-shaped mask — and the fixes were unglamorous: better chunking, a reranker, and an eval set we should have written first.
Read →There's no magic in the Model Context Protocol. It's JSON-RPC, a handful of verbs, and a discovery handshake. The interesting work is all in tool design — and that's where most teams underinvest.
Read →A walkthrough of how I decomposed a single overloaded chatbot into a coordinator and three specialist agents using Google's Agent Development Kit — including the part where I over-engineered it and had to delete an agent.
Read →Kubernetes assumes your pods are cheap, fast to start, and stateless. GPU inference pods are none of those things. Here's what that breaks — HPA, probes, rollouts — and how to work with the scheduler instead of against it.
Read →Every cluster I've inherited had the same handful of problems. None of them are exotic. All of them cause a 3am page eventually.
Read →Amazon Web Services · 01/2026 — 01/2029
Founder
Founded and lead a UK-based community of AI Engineers, Cloud Architects, and Full-Stack Developers building production-grade AI and cloud-native systems.
Open to conversations on platform engineering, GenAI architecture, and Kubernetes at scale. Reach out at ganeshs404@gmail.com or connect on LinkedIn.