Real-world AI applications took center stage at this year’s AFAC (Advanced Fintech AI Competition), where winning teams demonstrated technologies ranging from sign language translation and satellite imaging for credit risk to privacy-preserving bank fraud detection—all designed with global scalability in mind.
Organized by Ant Group alongside academic partners such as Peking University and Nanyang Technological University of Singapore, the annual AFAC competition has emerged as a launchpad for practical AI solutions ready for cross-border deployment. This year’s winners showcased strengths in technology innovation, social inclusion, and regulatory adaptation.
Swarm learning helps banks collaborate without sharing data.
Among the winners was Hong Kong-based A.I. Phoenix Technology, which received a third prize for a swarm learning system that enables banks to collaboratively train AI models without sharing sensitive data. The platform has already been deployed at financial institutions in Hong Kong for use cases including anti-money laundering and compliance checks.
“Regulators demand both data privacy and effective risk detection. Those two goals often conflict in traditional AI systems,” said Zhouming Xu, founder of A.I. Phoenix Technology and a former quantitative analyst. “Our system uses both federated learning and blockchain to allow model training across institutions without moving raw data and further protects model parameters simultaneously.” He emphasized the application’s relevance in cross-border settings, particularly in tightly regulated regions like Europe and Southeast Asia.
The team includes graduates from Harvard, Columbia, and Cambridge, as well as veterans from global firms such as HSBC and Amazon. Xu also acknowledged the hurdles small startups face. “Resource constraints force us to be highly efficient. We focus not only on technology but also on building scalable solutions.”
He also spoke highly of the competition’s mentorship, noting: “The guidance from the judges was extremely specific, detailed, and highly accurate. Our team has rich experience in competitions, and compared to pitch events we’ve attended in the Greater Bay Area, which also offered some mentoring and judge feedback, this was by far the most impressive and impactful. That was a major takeaway for us.”
Breaking barriers with real-time sign language AI
Another third prize went to Xiamen Blue Sky Beyond Technology, a student-led project developing an AI app that translates sign language into text in real time. The team began the project after volunteering at a special education school, where they saw communication barriers faced by people with hearing impairments.
“Many students with hearing impairments struggle with daily interactions, whether at the bank, hospital, or in public. We wanted to build something that helps include them more easily,” said Tianyi Li, founder of the startup. He described the challenges of working with limited data and computing power. “There are no open datasets for sign language. We collected and annotated all the videos ourselves.”
While the app is still in development, the team aims to distribute it free of charge to special education schools. “Supporting assistive technology isn’t always a priority. But we believe technology should serve everyone,” Li added.
Satellites and spectral imaging bring transparency to lending
The second prize was awarded to Xiamen Aohan Technology for integrating satellite remote sensing with financial risk management. Its system uses hyperspectral imaging and synthetic aperture radar to identify changes in landscapes and infrastructure, providing data useful for credit assessment and compliance in sectors like agriculture and natural resource management.
“Our satellites revisit locations every 15 to 30 minutes, making it possible to track changes at high frequency,” explained a team representative. The group includes PhDs from the University of Pennsylvania and professionals with experience from major aerospace institutes. “This is especially useful for auditing remote or cross-border projects where physical verification is difficult.”
The representative noted that while satellite data offers broad coverage, it works best alongside ground validation. “The technology isn’t for every institution, but it adds a valuable layer for large-scale or international operations.”
Now in its third year, AFAC is increasingly focused on solutions that address challenges across diverse markets – especially in emerging economies where fintech adoption is accelerating amid persistent regulatory and language barriers. What sets AFAC apart is not only its thematic focus but also its carefully designed interaction mechanisms that bridge innovative projects with industry expertise.
A prime example is the AFAC Meetup Time, which offered a structured yet personal platform for in-depth dialogue. Nearly 40 judges, including experts from Ant Group, NVIDIA, and Shanghai Hundsun JuYuan Data Service held one-on-one conversations with finalists, providing tailored advice on technology implementation, business models, and strategic development. This direct access to seasoned practitioners allowed participants to refine their solutions with precision – an opportunity especially valuable for early-stage startups.
Similarly, the dedicated Demo Time for the Startup Track enabled teams to present their prototypes to a broad audience of investors, corporate representatives, and technical experts. Unlike conventional pitch sessions, this segment emphasized collaborative problem-solving and resulted in concrete feedback on everything from architectural design to market entry strategies.
The outcomes were notable: this year’s competition attracted nearly 500 innovative proposals spanning wealth management, insurtech, low-altitude economy, Web3, and cross-border finance. The diversity of participants—ranging from university students and research labs to overseas returnees and seasoned entrepreneurs – underscored AFAC’s value as a global, interdisciplinary gathering for fintech innovation.
What united these projects was a shared focus on applying cutting-edge AI technologies – such as federated learning, long-context reasoning, and multimodal generation – to solve tangible problems in financial scenarios. Whether optimizing liquidity forecasting, validating multi-source data, or enhancing elderly financial services, these entries demonstrated both technical sophistication and practical relevance.
Through formats like Meetup and Demo Time, AFAC has strengthened its function as an innovation incubator – not only by selecting winning solutions but by facilitating genuine collaboration between rising talent and industry leaders. It is this emphasis on depth, dialogue, and real-world impact that continues to make AFAC a key driver of fintech innovation worldwide.
On September 10th，Winning teams have showcase their projects at the 2025 INCLUSION·Conference on the Bund in Shanghai this September, gaining exposure to international investors and financial institutions. As AI continues to evolve beyond hype into real-world applications, events like AFAC offer a glimpse into how technology can be built once and scaled everywhere – with empathy, compliance, and local nuance in mind.
