From Theory to Impact: NTU Students Excel at HOYA Chatbot Hackathon 2025
Team Untitled – comprising Cheng Yi Hsuen (Year 2, DSAI), Wang Shi Ying (Year 2, DSAI), Risha Sunil Shetty (Year 2, CSC), and Singh Janhavee (Year 2, CSC)—secured 1st Runner-Up at the NTU CAmpcOde x HOYA Chatbot Hackathon 2025.
Organised by the NTU Career & Attachment Office (CAO) in collaboration with HOYA Electronics, the industry-partnered hackathon challenged student teams to build an enterprise-ready HR chatbot to enhance internal communications and workflows.
Though still in their second year, Team Untitled approached the challenge with focus and curiosity. Coming off a previous hackathon, the team quickly realised the leap from theory to implementation was steep. Fortunately, they found direction through the Applied LLM & Microsoft Azure Workshop Series, organised by CDEFG & the Innovation Lab at NTU CCDS, and guided by Professor Ong Chin Ann.
The sessions, part of a six-part applied AI training programme, exposed the team to concepts in prompt engineering, retrieval-augmented generation (RAG), vector stores, and information retrieval. Inspired and equipped, the team went on to integrate external tools like Neo4j Graph and Computer Vision into their chatbot solution.
Stakeholders at HOYA were impressed—so much so that Team Untitled’s prototype was used as the benchmark for evaluating other submissions. According to the team, that moment was when it all came together.
“At the start of the year, none of us had experience building chatbots. We had to learn fast, fail forward, and rely on each other. The support we received from mentors and the NTU Innovation Lab made all the difference.”
More than a technical victory, this was a milestone in learning how to build with users in mind. The experience deepened the team’s appreciation for stakeholder engagement, agile development, and turning feedback into product decisions. For a group that started with just workshop concepts, it was a powerful reminder that impact happens outside the classroom.
