Work
Proof first, then the thinking behind it. Case studies from production engagements, and the notes we write along the way.
Case studies
ProofDesigned and deployed an AI agent pipeline that transformed a four-person development team from shipping 1-2 features per sprint to consistently delivering 11 - without adding headcount.
15 April 2026 · Software / SaaS
Read case study Read case study: Agentic development infrastructure for a software companyAn AI-powered system that detects, tracks, and classifies drones and other airborne objects in real time, protecting critical infrastructure with sub-second response latency.
12 March 2026 · Defence / Critical infrastructure
Read case study Read case study: Real-time drone detection for critical infrastructureTransformed a legacy defence computer vision solution from a monolithic, hard-to-maintain codebase into a modern, modular system built on DevOps and MLOps best practices - enabling faster feature development, stronger security, and reliable on-premise deployment.
1 March 2026 · Defence / Computer Vision
Read case study Read case study: Modernising a legacy computer vision solution for defenceA data pipeline that orchestrates thousands of monte-carlo style simulations per day - replacing manual runs with a fully automated, traceable system that produces the data foundation for real-world ML models.
1 December 2025 · Defence / Simulation
Read case study Read case study: Automated simulation pipeline for defence data foundationsEvaluated the security posture of an AKS-based Kubernetes platform for a payment processing client migrating from on-premise servers to Azure, assessing compliance against ISO 27001 controls.
15 October 2025 · Fintech / Critical infrastructure
Read case study Read case study: Securing a payment platform for ISO 27001Writing & notes
PerspectiveThe difference between AI-assisted development and agentic development is not incremental. It is architectural. Most teams are using AI wrong.
20 April 2026 · Peter Hinge · 7 min read
Read article Read article: Your team is the bottleneck, not your backlogMost ML predictions can run as a batch job. Knowing which ones actually need an API is the decision that determines your infrastructure cost, complexity, and reliability.
19 March 2026 · Mico Boje · 5 min read
Read article Read article: Not everything needs real-time inferenceThe same five problems show up in almost every production Kubernetes cluster we audit. Here is what they are and what to do about them.
12 January 2026 · Mico Boje · 7 min read
Read article Read article: What we find in most Kubernetes cluster reviews