Building software with agentic AI.

I started with GitHub Copilot on GPT models for inline assistance, then moved into editor-native agents with Cursor and Augment. Today my daily driver is Claude Code — driving real, multi-step engineering work across large codebases rather than just completing lines.

I've also gone hands-on with self-hosted models, running local inference with Ollama and vLLM to understand the trade-offs of owning the stack — cost, privacy, latency and control.

At Locate2u — a last-mile logistics and route-optimisation platform on a .NET back end and React front end — I set technical direction and architecture standards, and served as CTO through 2024. A core part of that leadership has been driving the team's adoption of AI-assisted and agentic development: establishing practices, tooling and review habits so the whole engineering organisation ships faster without sacrificing quality.

I stay deliberately hands-on. The same agentic workflows I ask the team to adopt are the ones I use every day to deliver features across fleet, routing and last-mile delivery.

I'm building a new platform designed for AI tools to build merchant and carrier apps — and, fittingly, the platform itself is developed entirely with Claude Code. It's both a product and a proof: a working demonstration of how far agentic development can go when you architect for it from the first commit.