A practitioner-academic building AI that earns its place in production.

Jon's PhD focused on novel neural network architectures for natural language understanding — work that seeded a long-running research programme on multi-recurrent networks (MRNs): deep recurrent ensembles whose recurrent state is structured to capture multi-scale temporal dependencies. MRN ensembles remain at the heart of Perceptronix's forecasting and classification systems today.

Perceptronix itself was founded in 2018, after Jon met a group of capital-markets investors at the OECD in Paris in 2015. What began as a research-driven response to the question “can modern neural ensembles outperform classical econometric models in real markets?” became a UK-registered consultancy that today serves clients across capital markets, sports analytics, and macroeconomic policy.

Across 20+ years he has shipped AI into production for trading desks, sports analytics teams, healthcare research groups, and cross-disciplinary economics collaborations. His current academic role with NIESR and the University of Birmingham produced MASCET, an EPSRC-funded macroeconomic forecasting platform whose neural ensemble has, on multiple occasions, outperformed the Survey of Professional Forecasters at detecting US inflation turning points by as much as five months.

Forecasting is only half the story. Alongside the modelling work, Jon applies statistical inference and experimental design — hypothesis testing, confidence intervals, causal identification, counterfactual analysis and A/B test evaluation — so clients can separate the effect of their decisions from background noise and act on uplift estimates with appropriate statistical confidence.

Beyond research, Jon led the MSc Data Analytics for Business at Nottingham Trent University as part of the school's Senior Management Group — a remit covering more than 120 programmes, 400 modules, and 200 academic staff. The MSc he designed went on to attract students from Credit Suisse, the NHS, HM Government, Equifax, and Rolls-Royce, and he has supervised six PhDs to completion.

What clients consistently say they value most: a no-nonsense, evidence-led approach that treats AI as engineering — versioned, monitored, validated, and accountable — rather than as a demo.