A 48-hour international AI sprint where teams created their own datasets, trained computer-vision models and turned them into working demonstrations.
EVENT OVERVIEW
01 / THE BRIEF
Perceptron AI Labs designed the edition around the complete AI-development lifecycle. Instead of beginning with ready-made labeled data, teams had to create or collect their own datasets before cleaning and annotating them. They then trained models against practical challenges in medical AI, sports analytics, satellite and remote sensing or open innovation.
02 / OUR APPROACH
The program combined pre-event onboarding for the Auta annotation platform with a residential build sprint, virtual participation for international teams, expert keynotes, hands-on mentorship, model testing and a final showcase. Teams moved through one connected pipeline: data collection, labeling, model training, inference testing, demo development, pitching and judging.
03 / WHAT WE HANDLED
04 / PROGRAM TIMELINE
Bachelor-level teams of two or three joined from institutions in Nepal and abroad.
A pre-event workshop prepared teams to collect, organize and annotate data using Perceptron's Auta platform.
Teams began without premade labeled datasets, gathering or creating the data required for their chosen problem.
Participants cleaned and labeled their datasets, trained computer-vision models and improved performance through testing.
Model inference was connected to a usable demo while mentors supported technical decisions and delivery.
Finalists demonstrated their systems and explained their impact to the judging panel.
Judges selected a champion and two runners-up from the final presentations, backed by a NPR 262,500 prize pool.
05 / PHOTOGRAPHY
06 / OUTCOMES
National and international teams took part through a residential Kathmandu sprint and a connected virtual format.
Every team moved from original data collection and annotation through model training, inference testing and a working demonstration.
Yaphongsan Yap, Zhan Quan and Loo Yi Sheng of Taylor's University won with FinchPoint, a computer-vision system for poultry-farm monitoring.
Rakshak Sigdel, Bipin Subedi and Prasun Bhattarai of Itahari International College secured first runner-up.
Nishan Maharjan, Pragyan Shrestha and Rojit Khadgi secured second runner-up.
The awards distributed NPR 150,000 to the champion, NPR 75,000 to first runner-up and NPR 37,500 to second runner-up.
NEXT CASE STUDY
READY WHEN YOU ARE
Share the institution, audience, expected format and the outcome you are aiming for. We will help turn the initial brief into a practical event plan.