Businesses are amassing more data than ever to fuel their AI ambitions, but at the same time they’re also concerned about who can access that data, which is often very private. PVML offers an interesting solution by combining a ChatGPT-like tool for data analysis with the security guarantees of differential privacy. Using recovery production augmented (RAG), PVML can access a company’s data without moving it, removing another security issue.
The Tel Aviv-based company recently announced that it has raised an $8 million round led by NFX, with participation from FJ Labs and Gefen Capital.
Image Credits: PVML
The company was founded by the husband and wife team Shachar Schnapp (CEO) and Rina Galperin (CTO). Schnapp earned his PhD in computer science, specializing in differential privacy, and then worked in computer vision at General Motors, while Galperin earned her master’s in computer science with a focus on artificial intelligence and natural language processing and worked on machine learning projects at Microsoft.
“A lot of our experience in this area has come from our work in large corporations and big companies, where we’ve seen things not be as efficient as we expected as naive students, maybe,” Galperin said. “The main value we want to bring to organizations as PVML is the democratization of data. This can only happen if, on the one hand, you protect this highly sensitive data, but, on the other hand, allow easy access to it, which today is synonymous with artificial intelligence. Everyone wants to analyze data using free text. It’s much easier, faster and more efficient — and our secret sauce, differential privacy, enables this integration very easily.”
Differential privacy it is far from a new idea. The basic idea is to ensure the privacy of individual users in large data sets and to provide mathematical guarantees for this. One of the most common ways to achieve this is to introduce a degree of randomness into the data set, but in a way that does not change the data analysis.
The team argues that today’s data access solutions are inefficient and create a lot of overhead. Often, for example, a lot of data must be removed in the process of allowing employees to gain secure access to the data — but this can be counterproductive because you may not be able to effectively use the logged data for some tasks (plus the additional time data access means that real-time use cases are often impossible).
Image Credits: PVML
The promise of differential privacy means that users of PVML do not need to make changes to the original data. This avoids almost all overhead and unlocks this information securely for AI use cases.
Basically everything large technical companies now use differential privacy in one form or another and make their tools and libraries available to developers. The PVML team argues that in fact it has not yet been put into practice by most of the data community.
“Current knowledge about differential privacy is more theoretical than practical,” Schnapp said. “We decided to take it from theory to practice. And that’s exactly what we’ve done: We develop practical algorithms that work best on data in real-world scenarios.”
None of the differential privacy work would matter if PVML’s real tools and data analytics platform weren’t useful. The most obvious use case here is being able to chat with your data, all with the guarantee that no sensitive information can be leaked in the chat. By using RAG, PVML can reduce ghosting to almost zero and the overhead is minimal since the data remains in place.
But there are other use cases. Schnapp and Galperin noted how different privacy also allows companies to now share data between business units. In addition, it may also allow some companies to monetize access to their data to third parties, for example.
“Today in the stock market, 70% of trading is done by AI,” said Gigi Levy-Weiss, general partner and co-founder of NFX. “This is a glimpse of things to come, and organizations that adopt AI today will be a step ahead tomorrow. But companies are afraid to link their data to AI because they fear exposure — and for good reason. PVML’s unique technology creates an invisible layer of protection and democratizes access to data, enabling revenue use cases today and paving the way for tomorrow.”
