Consultancy
Selective engagements.
Methodology review, validation strategy, and applied machine learning for research groups and small teams. Engagements that produce something durable, not a slide deck.
What i offer
Four kinds of engagement.
01 Methodology review
An external set of eyes on the design of a study, the choice of analytical methods, the validation strategy, and the interpretation of results. Useful for grant applications, manuscript preparation, and preregistration.
02 Applied machine learning
Building or improving a specific pipeline: classification, prediction, image analysis, sequence analysis. From problem framing through to deployment, with attention to the assumptions baked into each modelling choice.
03 Statistical advisory
Help thinking through what a dataset can and cannot support, where confidence is warranted, and where the experimental design forces caveats. Particularly useful when the data is small, the effects are subtle, or the stakes for getting it wrong are real.
04 Bespoke training
Sessions delivered to a specific team on a specific problem. Different from the public cohorts because the curriculum is shaped to what the team actually needs rather than to what reads well in a catalogue.
How an engagement works
From first conversation to final report.
01 Initial conversation
Free, no commitment. Twenty to thirty minutes to understand what you are working on and whether I can help. If I cannot, I will say so, and where possible suggest someone who can.
02 Written proposal
A short document setting out scope, deliverables, milestones, and price. Engagements typically run between two and twelve weeks depending on complexity.
03 The engagement
Working sessions, written work, and material delivered against the milestones in the proposal. Weekly review calls for engagements over four weeks.
04 Final report
A written report covering what was done, what was found, what is recommended, and what to be cautious about. Code, notebooks, and datasets handed over with documentation.
Scope
What I don’t take on.
Being explicit about what I decline to work on saves time on both sides. The list below is not exhaustive, but the pattern is consistent: I take on engagements where the problem is concrete, the data has structure, and the stakeholders are willing to act on findings even when those findings are inconvenient.
- Generic “AI strategy” or positioning work without a specific domain problem attached.
- Cryptocurrency, blockchain, or web3 projects.
- Pure data cleaning or labelling work where no analysis follows.
- Engagements where a conclusion has already been decided and the consultant is being hired to validate it.
- Work that conflicts with my University of Malta affiliation or with another active engagement.
Get in touch
Alternatively
Live cohorts and self paced courses.
For groups and individuals who want training on applied machine learning rather than a bespoke engagement.