Collaborate
Research, speaking and community.
My day job is building data and AI systems at AI Creative Solutions. Alongside it I research, write, mentor and speak. If you are planning a study, a story or an event, I would like to hear about it.
01 Ways to collaborate
Four ways to start.
Research collaboration
Models that hold up outside the lab: domain shift, scarce labels, field conditions and Earth observation.
First step: Share the question and the data you can access.
Interviews, writing and media
Comment on AI, data skills and higher education in Africa.
First step: Send the outlet, the angle and the deadline.
Mentoring and community
Capstone mentoring, Zindi challenges and community events in Zimbabwe.
First step: Tell me where you are stuck.
Speaking and workshops
Keynotes, panels, guest lectures and hands-on workshops on data engineering, applied machine learning and governed AI, for technical and business audiences.
First step: Send the audience, the date and the format.
02 Talks
Talks I give.
Data and AI in business
From ERP tables to board packs
Data engineering that earns trust: pipelines, lakes and warehouses built on undocumented ERP data, and the reconciliation that proves the numbers before anyone acts on them.
Governed AI agents on live business data
What it takes for an assistant to answer and act on production data without exceeding the authority of the person using it.
Where data science creates business value
Choosing problems where a model, a dashboard or a pipeline changes a decision, and measuring whether it did.
Research
AI that holds up outside the lab
Why models that excel on benchmarks fail in the field, and how to evaluate for low light, scarce labels, basic devices and intermittent power.
Skills and leadership
Data skills as the new literacy
What students and professionals across Africa need to learn now, and how competitions and communities such as Zindi build those skills.
Leading without a title
Building authority through work, not position.
Formats: keynote, panel, workshop, guest lecture.
03 Approach
How I work.
Evidence over claims.
If a number cannot be verified, it does not go on the page.
The model never does the arithmetic.
Queries and code calculate. The model plans and explains.
Build for the terrain.
Budget phones, intermittent power, multi-currency ledgers. A system that only works in ideal conditions is a lab result.
Honest status beats fake completeness.
Unknown renders as unknown, never as a fabricated zero.