About javai.org
javai.org is the technical home of the Javai project family — open-source frameworks and tooling for testing systems whose behaviour is non-deterministic by nature: LLM-backed services, recommender systems, ML pipelines, and any service whose output is a distribution, not a single value.
The site hosts the projects, explains the methodology behind them, and publishes Signals — long-form essays on software engineering, software quality, and engineering leadership in an age when AI has reset both the economics of writing code and the epistemics of trusting it. Signals draws on decades of practice and on recent work with machine learning, AI, and agentic systems. It is written for serious engineers, the leaders responsible for them, and the investors whose money is exposed to the difference.
The companion site javai.ch covers the regulatory side — AI compliance developments affecting Switzerland and the EU.
Founder
Mike Mannion is a senior software engineer and consultant with more than thirty-five years of experience building enterprise software, data platforms, and — more recently — AI-native systems. His work focuses on the question that makes AI hard to ship and harder to trust: not whether the system runs, but whether its behaviour is reliable, interpretable, observable, and economically sustainable.
Mike created punit, the probabilistic testing framework for Java, and is the founder of both javai.org and javai.ch. He advises investors and boards on the technical and operational credibility of AI-enabled software companies — architecture, model-risk exposure, testing maturity, observability, and long-term maintainability — distinguishing companies whose technology supports their valuation from those whose demos conceal fragility.
He writes and speaks on probabilistic testing, AI-native engineering practice, and evidence-based software quality.
Get in touch
Questions, suggestions, corrections, or a project to discuss?