Ad Blocker Detected
Our website is made possible by displaying online advertisements to our visitors. Please consider supporting us by disabling your ad blocker.
“Data is food for AI,” observed Andrew Ng, the artificial intelligence guru who co-founded Google Brain and led Baidu’s AI group. However, there’s a problem implied by the analogy. Artificial intelligence systems do indeed need data to function effectively, to learn and grow, and to improve. But as we all know, the best food takes far longer to prepare than to consume; and so it is with data.
That’s where Zurich-based Lightly comes in. The Swiss company, which is today announcing the successful completion of a $3 million seed funding round, believes its technology can take much of the hard work out of preparing data so that it can be served up to AI and machine learning models. The aim is to ensure these models eat exactly the right data for their needs – and therefore to generate better outcomes.
“Any machine learning model is only as good as the data it is trained on,” points out Matthias Heller, co-founder of Lightly. “We’re helping the model to select really high-quality data.”
In practice, Lightly’s technology achieves this goal in two different ways. First, it can identify the data that will help the model learn most effectively – for example, by screening out new data that is very similar to information the model already holds in large quantities in order to focus on the data that brings new value. Second, it automates the process of labelling the data in order to ensure the model can put it to the best possible use.
“We deploy self-learning algorithms that tell data teams which 1% of their data is the most valuable,” summarises Heller. “With our approach, companies see their labelling costs decrease by up to 90%, while their AI model can improve by 20%.”
Lightly founders Matthias Heller and Igor Susmelj
Lightly sees its technology being applied in a wide array of environments. The autonomous vehicle sector is one obvious application, given the vast amount of data that self-driving cars need in order to navigate safe passage. Medical imaging is another promising area, with doctors increasingly using AI to analyse images they take when trying to diagnose disease.
Importantly, explains Heller, Lightly’s skill lies not in any one application of AI and machine learning, but in the processing of the data required. “We’re not trying to tell engineers how to do their jobs, but we are giving them the tools to make it easier to do those jobs,” he says. “If you’re a medical imaging technician, you shouldn’t be wasting your precious time trying to build the best possible data infrastructure, any more than we should be trying to interpret medical images.”
The company also hopes its technology will address some common problems that have dogged the AI sector. For example, with better selection of data, rather than simply adding more of the information, it may be possible to reduce the risk of bias creeping into AI models.
Founded in 2020 by Heller and his business partner Igor Susmeli, Lightly appears to have struck a chord in the industry. Its open-source framework has attracting more than 2,000 stars from developers on the GitHub platform, and paying customers are following. “We’re seeing real traction and onboarding new customers every month,” Heller adds.
The company believes it can now accelerate its growth with a concerted focus on the US market, and has plans to open an office in Silicon Valley in the coming months. Those expansion efforts will be boosted by the $3m of capital it has raised in a seed round led by Wingman Ventures. The funding is earmarked for further investment in the product, as well as expansion of its US footprint.