“Ambition should wake you up every morning instead of an alarm.”
Founded in: 2022
While pursuing a Ph.D. in Power Electronics and Artificial Intelligence (AI), Fannie Lin learned about the benefits of AI’s applications to make life easier, smarter, and safer. She got firsthand experience of the process end to end; from ideation to conceptualization of use cases to development and refinement of the AI model to the result of the application. Through this journey, Fannie recognized that the current AI environment is capital and time intensive and thus, inaccessible to many people and enterprises.
“I believe in the power of technology, and particularly in AI, that has both social and economic impact. But it should not just stay inside a laboratory or be limited to the field of research, it should be diffused to every aspect of our real lives. And I realised that entrepreneurship can make this possible.”
This conviction led Fannie to launch Ailiverse, which is developing a software solution to make deep learning (beginning with computer vision), accessible to all enterprises regardless of size and industry. The solution empowers companies to build computer vision models with incredibly less training data and in a shorter time than what is required by the conventional approach. Traditionally, 50,000 to 100,000 or more data points – images and video footage – are required for training a model from scratch. Collecting and annotating these large data sets is time-consuming and can take up to several weeks or months. In the subsequent stage, GPU training, which involves building a neural network based on this data set, takes several more weeks. Collectively, the long developmental pipeline drives up the overall cost exponentially.
Ailiverse, which has developed breakthrough technology in the nascent space of unsupervised domain adaptation, along with few-shot learning, enables companies to build computer vision models with just 10% of the traditionally required data. It makes data collection fast and cost-efficient without compromising on accuracy. Furthermore, Ailiverse makes the neural network so small that it accelerates the GPU training time by 10 times. Thus, instead of weeks, training an AI model takes just a couple of minutes or hours and it is then ready for deployment and testing.
“ChatGPT, which has taken the world by a storm today, is trained on billions of data points that require expensive servers and a very long pipeline. A single exercise of training from scratch can cost nearly US$750,000. Our vision is to transform and democratise AI to make it accessible to all enterprises.”
Fannie and co-founders Burhan Ul Tayyab and Kelvin Kok believe that there’s an urgent need for AI solutions like Ailiverse to meet the growing demands of companies that are getting more and more aware of AI’s benefits. Companies across industries – manufacturing，retail, security and defence – are recognising how this technology can empower them to enhance and transform their businesses. What motivates Fannie is that making AI more accessible will help startups and SMEs integrate deep tech solutions into their business models to solve large-scale problems.
In a short span since its inception, Singapore-based Ailiverse has customers across sectors including defence, security, retail, and logistics. Before co-founding Ailiverse, Fannie launched an education startup to provide affordable English language education to Chinese university students.