Work / Experience & technical practice

Experience behind
the advice.

From early AI prototypes at Camunda and the Semaverse.ai research platform to an international consulting practice of more than 125 people. My work spans the strategy, the organisation and the engineering needed to deliver.

Read the career context and approach

Selected leadership & advisory work

The remit.
The work behind it.

Semaverse.ai · Venture work

Connect research, evidence and working documents.

Co-founder; platform architecture and engineering

Designed and built an AI research platform for M&A and market intelligence: research agents, persistent memory, knowledge graphs and cross-source search. The platform combined web crawling and document ingestion with tenant isolation and fine-grained access controls. A two-way PowerPoint workflow turned corporate templates into research pipelines and the results into branded decks.

Semaverse became a PitchBook Premium Connector partner and a Crunchbase MCP launch partner. My partnership and integration work also included Tracxn, and helping Crunchbase work through MCP and OAuth implementation details. The engineering responsibility extended from source access and permissions to the research and documents people worked with.

Camunda · Consulting through Bold Data

Give an emerging AI strategy something to test.

Early AI product and technology advisory

Helped establish early AI product direction at the process-orchestration company, building prototypes that informed its product and technology strategy. They were used internally and subsequently demonstrated at Camunda conferences, giving product and engineering teams something concrete to evaluate and build on.

The work included a documentation and code assistant that accounted for different software versions, and a prototype assistant for Camunda’s expression language. For the latter, I used a larger language model to generate thousands of training and test examples, then fine-tuned a smaller model to help users write and correct expressions.

Holland & Barrett · 2019–2021

Turn external spend into internal capability.

Director of Engineering, Data and Analytics (CDO)

Redirected third-party budget into an in-house data and analytics capability. Built data engineering, business intelligence and data science teams, alongside a cloud-native platform connecting legacy systems and a growing event-driven architecture.

The capability supported reporting, forecasting, supply-chain decisions, purchasing and range planning.

Think Big, a Teradata company · 2016–2017

Lead delivery across four continents.

Partner, International Engineering Practice

Led an international practice of more than 125 people across data and software engineering, systems engineering and solution architecture. The remit covered enterprise and public-sector delivery across four continents and fifteen time zones.

Responsibility included practice growth, reorganisation and integration within Teradata.

Worldpay · 2018

Connect strategy and the operating model.

VP Data Strategy (interim)

Worked with the CDO on data, platform and analytics strategy in a global business. The scope included data science, machine learning, AI and the operating model needed to support them.

An interim strategy remit: making the technology choices and how the organisation works part of the same discussion.

Early Bold Data / Independent engineering

250 million product extractions.
Built single-handedly.

After Holland & Barrett, I built a scale-out crawler and scraper running across Google Cloud and AWS. It collected approximately 250 million Amazon product extractions across the US, UK, Germany and other markets. That is a count of extractions, not unique products.

The technical achievement was collecting at that scale with low operating costs, as a solo build. I did not turn it into a commercially viable business. It is a useful distinction: an efficient platform and a viable market are two different things to establish.

The published datasets and analyses are part of that work, not the full collection.

These are historical releases, not a promise of a current data feed. Check the collection dates and licence before use.

Explore the published dataset

The current workshop

Three projects.
Three human bottlenecks.

Development snapshot · 6 September 2026. These are independent projects, not client case studies or evidence of commercial outcomes. They are still being reviewed or developed and are not ready for general use.

Human review

Tax Shrink

Rules & transparency

Exploring salary, dividends and tax scenarios for UK company owner-operators.

Next human step

Review the calculations, assumptions and generated copy before wider promotion.

MVP in development

Llamar

Knowledge & context

A browser extension exploring how to collect and work with useful information from the web.

Next human step

Stabilise one useful workflow, the extension and its website for a first public version.

Playtesting & curation

Turmberg

Systems & product judgment

A strategy game testing where generated systems and content meet human ideas of what is fun.

Next human step

Play, simplify and tune the levels, units and mechanics. More generated content is not the next step.

What comes out of the work

Decisions worth sharing.

The useful story includes what failed, what was removed and what needed human judgment.

I write about the economics, architecture and verification behind the work. A finished feature is only one kind of result. Finding the wrong assumption early can be more valuable.

Read the insights

Start with the decision

What is at stake
for your business?

Tell me the decision you face, its timing, and what happens if it goes wrong. We can establish whether my involvement would be useful.

Discuss the decision