About Me

I'm Paul Kohler - a software architect in Melbourne who has been shipping systems since 1997. These days I'm a Solution Architect on SEEK's Data & AI team, working on the generative AI capabilities behind one of Australia's biggest online marketplaces: job-candidate matching, assistants, hirer automation, and the architecture patterns and guardrails that help teams ship AI features safely.

I've been at SEEK for over a decade, and I've had most of the jobs: Principal Developer on the core Ad Centre, helped kick off the Online Identity team (unified employer login on Auth0), Technical Lead growing Candidate Management to 14 developers while rebuilding the platform on an event-driven architecture, Performance Engineering lead keeping the platform fast through a run of acquisitions, Staff Engineer, and now architect. That range is the point - I've owned the code, the team, the incident, and the diagram, so my architecture comes with production scars.

Before SEEK: consulting, government, policing systems, banking middleware, and a university web gig back when ASP classic was a reasonable choice. Twenty-five-plus years of the industry reinventing itself is good training for the current one - I've watched enough hype cycles to know the difference between a demo and a system, which is exactly the skill the AI era rewards.

The AI work isn't just my day job. Under my own AI Creations banner I build and ship products: Agentic Diagrams, a tool for visualising agentic systems with an AI Architect that designs alongside you - built on GCP with multiple LLM providers, and anchored by an open-source schema (.agentic.yaml) so agent architectures can live in version control instead of screenshots. And Kodr, a zero-dependency coding harness for local LLMs that I'm building as a learning exercise - every phase documented on the blog, failures included, with evals to keep me honest.

How I work: clear architecture with minimal complexity, small increments shipped often, strong defaults for observability, security, and reliability. I see the value in deleting components rather than adding continuously, and I think the best architecture document is one an agent can read too.

If you want to talk architecture, agentic AI, AWS/GCP, or TypeScript systems - find me on LinkedIn or GitHub.