Creating AI Platforms
What I do
Twenty years of engineering, and the last few spent where the complexity actually is. GPU workloads, LLM serving, and the platforms teams ship on every day. Not proofs of concept, and not demos that fall over the moment someone depends on them.
01
Entrepreneur
I co-founded techmeout.io in 2021 and ran it for close to four years, building GenAI integrations for enterprise systems and the microservices architecture around them. Now I'm building Sciensoft Technologies. Starting something means owning the parts nobody hands you, and that is the part I want.
02
AI Platform Engineer
At TKH Artificial Intelligence I lead the platform an entire business runs on: model serving and GPU work sit beside the applications, workflows and databases that pay for them, on one hybrid Kubernetes estate across Azure and on-premises, networked and provisioned entirely as code. Deploying the tools is the easy half. The rest is psychological safety, clear ownership, and security a team owns from the start rather than inherits at the end.
03
Creator
I make the tools I needed and could not find, and I write about the problems I've had to solve myself. I play guitar, I ride motorcycles, and I build things that did not exist before. The through-line is craft, and I don't separate the technical from the personal.
04
Where I've Built It
TKH Artificial Intelligence · now
A hybrid Kubernetes estate across Azure and on-premises Proxmox: LLM serving with vLLM and LiteLLM, GPU fine-tuning for the ML teams, and the applications, search and databases the business depends on. The Azure side is hub and spoke: VNets and subnets behind NSGs and private DNS, Front Door and a WAF at the edge, Key Vault for secrets, and every product reached through one API gateway with its own identity and route per customer. All of it is Terraform, and workloads that sit idle scale to zero. I own the technical roadmap through 2027, with security and cost reduction at its centre, and the ISO 27001 preparation.
techmeout.io
GenAI integrations, microservices, and the company around them. Integration architecture with LangChain and prompt engineering, running on GCP and Kubernetes.
Getir
AWS platform supporting 500+ developers. Networking, databases, Kubernetes and GitOps with ArgoCD and Terraform, with shift-left security throughout.
IKEA
Two teams through IKEA's digital transformation. A greenfield carbon footprint tracker across 8+ countries on 100% infrastructure as code, and a product live in 30+ countries and 349+ stores.
Swapfiets
Monolith to microservices using Domain-Driven Design, Event Sourcing and CQRS. Azure App Services to AKS, with Helm, canary deployments and zero-downtime releases.
Dell Technologies
Pioneer Investments
DoMore!
05
What I Build
I build with agents, and I keep the conversation. Every commit in these repositories has the prompt that produced it, and the decisions sit next to the code as written records, so the reasoning is still there when the code changes.
agentic-bootstrap
A persistent operating model for AI-assisted engineering. One markdown file installs the substrate, and every agent that opens the repository inherits it. This site is built under it.
mcp-connectors
Claude on the web and on mobile, reaching files and running commands on your own machine through a secure MCP connector. Social sign-in and a desktop tray app around it.
musical
Upload a track and ask it anything. Key, tempo, structure, the line that comes before the chorus. A multi-tool agent answers from the recording itself, which is the guitar half of me pointed at the AI half.