Build the dataset · Train the model · Evaluate the impact
IIMS x Perceptron International Hackathon 2026 is an intense 48-hour build sprint dedicated to bridging academic innovation and industrial application.
Unlike traditional software hackathons, our sprint highlights the complete AI development lifecycle. Rather than downloading premade labeled datasets, teams will construct, label, and deploy original datasets using advanced tools during the sprint.
This event aligns with the grand vision of IIMS 2.0 — next-generation academic transformation at IIMS College. Featuring upgraded smart learning spaces, collaborative code stations, and hands-on guidance, we aim to prepare tech talent for tomorrow's challenges.
Using Auta, Perceptron AI Labs' dataset creation platform, teams will organize, annotate, and clean their data from scratch. Tokens are initially capped to 10k images per user, allowing teams to build high-quality visual data segments.
Gathering raw images, video feeds, or tabular records responsibly from public domains or custom assets.
Organizing, tagging bounding boxes, segmenting classes, and cleaning dataset elements using Auta platform.
Training neural networks (YOLO, PyTorch, TensorFlow) using the newly annotated and exported custom dataset.
Running checks on raw unseen test samples to measure precision, recall, confusion matrices, and inference speed.
Developing lightweight interface wrappers (using Streamlit, Gradio, React, or FastAPI) to display model outputs.
Presenting workflow, dataset properties, training metrics, inference demos, and economic impact to judges.
Select a challenge path or chart your own route to design actionable artificial intelligence solutions.
Implement computer vision or prediction algorithms for diagnostics, cellular tracking, clinical support, or digital pathology logs.
Model player kinematics, tactical patterns, automated refereeing, ball path tracking, or performance metrics using video files.
Develop object classification or segmentation algorithms for agricultural patterns, disaster tracking, or urban changes using remote sensing.
Propose an original problem statement, assemble data targets, label items in Auta, and build custom AI prototypes tailored to societal impacts.
Track important dates from general registrations through to the live 48-hour build and post-event announcement.
Team registrations open internationally. Registered teams receive sample problem blueprints and introductory guidelines.
All candidate details lock. Registration fees (Rs. 5,000 per person) and document verification must be completed by midnight.
Orientation sessions with dataset-first models. Users receive active credentials and introductory training tokens.
48-hour continuous build at Aranya Boutique Hotel. Data sourcing, labeling, model optimization, and demo assembly.
Hosted at IIMS College campus. Selected teams demonstrate functional prototypes. Evaluated by international panels.
NPR 2,62,500 total cash awards to support tomorrow's machine learning projects, startups, and developers.
Supporting innovation, emerging technologies, and developer communities.
Review event workspaces. Note that the 48-hour build phase is a residential stay.
Nagpokhari, Kathmandu, Nepal
Residential Sprint. Dormitories and sleeping areas are provided by the organizers. Teams must remain at the facility. Hot meals provided.
Naxal-Bhagwatibahal, Kathmandu, Nepal
Final project evaluation, live pitches, judging, and awarding ceremony at the smart seminar halls of the newly launched campus.
