AUST team '4Bytes' wins international hackathon in Manila

A team of students from Bangladesh's Ahsanullah University of Science and Technology (AUST), named '4Bytes', has won first place in the 'Financial Health Frontiers' competition at the 2026 APRU Data, AI & Policy APAC Hackathon, held in Manila, Philippines.

The final round of the international hackathon took place from November 4 to 6 at the Asian Institute of Management (AIM) in Manila. From 162 applications submitted by more than 600 students from 15 countries and regions across Asia-Pacific, only 12 finalist teams from 10 countries and regions reached the final in Manila. The AUST team emerged champion from that field of 12. Every member taking part in the final round also received a travel grant of 800 US dollars from the organisers.

Dividing the work by skill, Jamila Mohammad took charge of product lead and modelling, Humayra handled data, Asmita looked after policy and the decision framework, and Mufi Al Dosari was responsible for development.

While the team was preparing to travel to Manila, their university's semester final examinations were under way back home. Caught between exams and the chance to represent the country on a world stage, they were backed by classmates and teachers, and their exams were postponed for their convenience.

The '4Bytes' project is called 'CredLens', an AI-assisted credit decision support platform. Much of the trading of the country's countless small and marginal business owners is recorded on handwritten receipts or scraps of paper; with no formal banking or credit history, banks and financial institutions cannot assess their creditworthiness when they apply for loans. The project was built to remove that problem.

Built on the real working methods and feedback of ShopUp, a leading B2B commerce platform, the prototype can analyse small traders' purchase, delivery and invoice data to automatically generate a reliable credit score and loan limit.

What impressed the judges most was the project's ethical and human side. Reading photographs of handwritten receipts with AI or OCR technology carries a risk of errors from unclear handwriting or numbers. CredLens does not blindly trust AI: it includes a 'Records Desk' that verifies any unclear information against the original receipt instead of guessing. Through a 'Merchant Profile', traders can see all their own information and apply for corrections if anything is wrong.

'4Bytes' beat teams from institutions including the Korea Advanced Institute of Science and Technology (KAIST), Seoul National University, the National University of Singapore and the University of Melbourne to take the title.

Expressing the feeling of victory, team member Asmita Guha Thakurta said: 'Representing the country at the international level and winning first place is truly hard to express in words. We succeeded on this journey not through the efforts of just four people, but with the support of a larger family made up of our teachers, classmates and families.'

The project currently runs on synthetic data, but the team does not want to leave it as just a hackathon project. Their future plan is to field-test it with ShopUp's credit officers.