Roystian Industry Challenge
Open-ended problems outside the structured curriculum. Challenge results feed your Career Readiness Score and are considered for the advanced internship pathway.
Six challenges
Pick the problem that fits your field.
Each challenge states its brief, its requirements and exactly what it will be judged on before you start.
AI Challenge
Build a model that makes a decision someone would actually act on.
You are given a dataset and a decision that currently gets made on instinct. Build something that improves on instinct, prove that it does, and state honestly where it should not be trusted. A model that is 95% accurate and unusable scores lower than one that is 80% accurate and deployable.
Judged on
- Problem framing — is the right question being answered?
- Technical execution — is the method sound?
- Honest evaluation — are limitations surfaced rather than hidden?
- Decision usefulness — could someone act on this?
Most relevant to: AI / ML · Data Analytics · Python
Coding Challenge
Ship a working feature against a specification written by someone else.
You receive a specification with three deliberate ambiguities and a codebase with an established style. Build the feature, resolve the ambiguities by asking the right questions in your write-up, and leave the codebase better than you found it.
Judged on
- Correctness — does it do what was asked?
- Ambiguity handling — were the gaps spotted and resolved sensibly?
- Code quality — would a reviewer approve this?
- Communication — is the reasoning legible?
Most relevant to: Full Stack · Python · Java · Software Dev
Data Challenge
Find the story in a dataset that nobody has looked at properly.
You are handed an operational dataset and a vague question from a stakeholder. Sharpen the question, answer it with evidence, and deliver a recommendation the stakeholder can act on this week.
Judged on
- Question framing — was the real question found?
- Analytical rigour — do the conclusions follow from the data?
- Data honesty — are quality issues disclosed?
- Actionability — could the stakeholder act on Monday?
Most relevant to: Data Analytics · Business Analytics · AI / ML
Software Testing Challenge
Break a build that its authors believe is finished.
You are given a build the team considers release-ready and its existing test suite. Find what the suite misses. Every defect you report must be reproducible by someone who has never seen the application.
Judged on
- Coverage analysis — were the gaps found systematically?
- Defect quality — are reports reproducible and precise?
- Prioritisation — does severity reflect user impact?
- Verdict — is the release call defensible?
Most relevant to: Software Testing · Software Dev
Business Case Challenge
Make a recommendation you would defend to a board.
A company faces a decision with incomplete information and real money attached. Build the case, state your assumptions openly, and be clear about what would change your mind.
Judged on
- Framing — are criteria set independently of the answer?
- Analytical quality — is the model sound and transparent?
- Assumption testing — is uncertainty handled honestly?
- Conviction — is there a real recommendation, not a hedge?
Most relevant to: Business Analytics · Sales & Marketing · Data Analytics
Sustainability Challenge
Separate a real environmental gain from a claimed one.
An organisation has published an impact claim. You have the underlying data. Determine whether the claim holds, what the defensible figure is, and what should be measured going forward.
Judged on
- Verification rigour — were confounders handled?
- Attribution — is the effect correctly isolated?
- Honesty — is uncertainty stated plainly?
- Forward design — is the recommendation practical?
Most relevant to: Sustainability · Mechanical · IoT & Electronics
Challenge entries are evaluated and scored. A score of 70% or above earns the Industry Challenge badge.