01
Security
Find and mitigate vulnerabilities in AI systems — prompt injection, data poisoning and adversarial inputs that exploit weaknesses in multilingual models.
Oct 2026 / Melbourne · Saigon South · Hanoi
RMIT Hackathon 2026
Security. Generative AI. Low-resource languages. Two days building solutions to the problem of GenAI inaccuracy in the world's underserved languages.
Registration Soon
Explore Themes
Hosted by
[ Themes ]
Three fronts,
Three fronts,
one problem.
Generative AI performs well in English and a handful of other well-resourced languages. Everywhere else it degrades — quietly, and often unsafely. This year's hackathon sits at the intersection of three areas.
02
Generative AI
Work directly with large language models to improve accuracy, reliability and grounding, particularly where current models are demonstrably weak.
03
Low-Resource Languages
Tackle GenAI inaccuracies in languages with limited training data, building approaches that make these systems more trustworthy for the communities who use them.
[ The Challenge ]
Solve what
Solve what
gets overlooked.
Teams will work on the problem of GenAI inaccuracies in low-resource languages: the failure modes that appear when a model is asked to operate in a language it was barely trained on. Hallucination rates climb, safety guardrails weaken, and evaluation tooling that works in English stops being reliable.
You might approach this through better evaluation, data augmentation, retrieval grounding, fine-tuning strategies, adversarial testing, or something we haven't thought of. The framing is deliberately broad because the interesting work happens at the edges.
The specific challenge brief is still being defined and will be published ahead of the event.
[ At a Glance ]
The essentials.
Oct2026 — dates TBC
When
03Melbourne · Saigon South · Hanoi
Campuses
04students per team
Team Size
02days, in person
Duration
[ How to Participate ]
Your route in.
Step 01
Watch for registration
Registration opens closer to the event. Details will be posted to the news page.
Step 02
Form your team
Four students from the same university, any mix of undergraduate, postgraduate or PhD.
Step 03
Build across two days
Work on the challenge in person at your campus, with mentorship from industry and university experts.
Step 04
Present your solution
Pitch to the judging panel and answer questions on your approach.
[ Who Should Join ]
Built for people who
Built for people who
want the hard version.
- RMIT University students — undergraduate, graduate and PhD — across Melbourne, Saigon South and Hanoi campuses.
- Students from other universities.
Teams of four students. Same university per team.
[ What You Get ]
Beyond the leaderboard.
01
Real technical depth
Hands-on work across AI, NLP, security and language technology — not a toy problem.
02
Expert mentorship
Direct access to practitioners working in AI, linguistics and security.
03
Work that matters
Contribute to making AI usable and safe for underserved language communities.
04
A wider network
Meet collaborators, mentors and employers across three campuses and two countries.
Two days. One under-served problem.
Registration Soon
Read the FAQ