Context
The Nedbank Machine Learning & Data Science Challenge 2026 is a national competition that puts entrants against a real predictive problem, judged on measurable model performance rather than opinion, exactly the evidence-led principle Nullius is built on.
The problem
The challenge required taking raw, imperfect data and producing a model that performed under objective evaluation, competing against 251 entrants including established teams.
Approach
We worked the problem end to end: exploring and cleaning the data, framing the target carefully, engineering features, and iterating across models with honest, held-out evaluation rather than chasing leaderboard noise. Decisions were driven by what the validation data actually showed.
Tech stack
Outcome
A 3rd-place finish out of 251 entrants (top 1.2%), demonstrating the ability to take a real, messy data problem and produce a model that performs under objective scrutiny.
Published June 2026