AviviD.ai and a National Taiwan University Team Join Forces to Innovate

MTSTRec, a multimodal recommender system jointly developed by AviviD.ai and an interdisciplinary data science and signal processing research team at National Taiwan University (NTU), was recently accepted as a paper at ICML 2025 (International Conference on Machine Learning), one of the world's top machine learning conferences, bringing the work onto the international stage. The research breaks new ground by combining diverse data such as product images, text, and prices and aligning them in depth along the time series, which significantly improves the accuracy and timeliness of recommendations on e-commerce platforms and marks a new breakthrough in applying AI to industry.

NTU Professor Che Lin said, "AviviD.ai's first-party data was a key driver behind the creation of MTSTRec. Academic research often has theories and models but lacks real, large-scale multimodal sequential data. AviviD.ai not only provided complete data but also allowed the team to validate and innovate in a real business setting, which let MTSTRec go from an academic idea to a technology that solves real pain points in the industry."

To meet the industry's need for "real-time dynamic recommendations," the paper proposes a time-aligned shared token (Shared Token with Time Alignment) mechanism that effectively integrates changes in multimodal data over time. This lets the recommender system respond quickly to user behavior, so personalized recommendations no longer lag behind. "Traditional methods tend to 'see a lot but miss the mark,' whereas MTSTRec can capture users' latest preferences in real time, improving conversion rates and business results," Professor Lin added.

Professor Che Lin also pointed out that Taiwanese companies such as AviviD.ai, which have built deep roots in the local market, can not only bring real industry problems and challenges to the AI research community but also work with academia to define scientific questions and design innovative solutions. "This kind of dialogue between industry and academia is a catalyst for putting theory into practice, and it is the collaboration experience we value most."

Going forward, MTSTRec aims to add features such as recommendation explanations, real-time user feedback, and A/B testing, continuing to evolve into an intelligent recommendation engine and extending further into predictive marketing. Professor Che Lin said, "This achievement shows that Taiwanese software companies have world-class competitiveness in AI recommendation. I hope AviviD.ai can enter overseas SaaS markets backed by academic support, technology recognized at a top conference, and real-world validation, setting an important milestone for Taiwanese AI SaaS companies expanding abroad."

AviviD.ai and NTU will also continue to deepen their industry-academia collaboration, promoting student internships, industry projects, government R&D programs, and more, to build Taiwan's AI ecosystem together. "Companies are not just sponsors but also coaches; students are not just apprentices but also innovation partners. We hope to set up a joint lab in the future so that talent and technology can advance together."

(From left) Cheng-Yu Huang, Business Development Manager, Data Expansion Department, AviviD.ai; Professor Che Lin, National Taiwan University

 

Image: AviviD

Finally, Professor Che Lin summed up AviviD.ai in one sentence: "AviviD.ai is the best 'Transformer' for bringing industry problems onto the world stage, turning Taiwan's real business challenges into goals the global AI community works on together."

To learn more about AviviD, please visit the official website for more information.