A smart use of AI
Paulina Grnarova strolls through the garishly bright rooms of the WOW Museum in Zurich, fascinated by the optical illusions around her. They remind the 30-year-old CEO of her childhood, when her fondness for geometrical complexity first developed. Her father, a computer science professor and entrepreneur, nurtured her interest in technology and mathematics from an early age and came up with fun ways to bring the material to life. “As a kid, I saw maths problems as puzzles to be solved. By the time I was seven, I was already entering maths competitions in North Macedonia, where I grew up,” says Grnarova. A place to study computer science at ETH followed in due course, and she eventually went on to found a software company based around artificial intelligence, or AI.
Grnarova loves the way the WOW Museum plunges visitors headlong into a world of mathematical and optical conundrums. She’s also a big fan of gamification in apps such as Duolingo, which she’s been using for years to improve her German.
“I’m a classic overachiever, not just at work but also in my personal life,” says Grnarova. She makes time for body combat, fitness training or dance lessons on an almost daily basis. But she admits it can be hard to switch off from work, and she often spends her free time reading books or listening to podcasts by other successful female entrepreneurs. Grnarova has always been driven to succeed: having graduated at the top of her class from the University of Skopje, she received a prestigious prize from the president of what is now North Macedonia, which gave her the opportunity to study abroad. She initially attended EPFL in Lausanne, before moving to ETH Zurich for her doctoral studies.
Artificial Intelligence for lawyers
Just over a year ago, Grnarova founded the company DeepJudge together with three other doctoral students from the Data Analytics Lab, which is led by Professor Thomas Hofmann. “Our platform is designed to revolutionise the way lawyers and other legal professionals work,” she says. AI-powered software takes over manual and other time-consuming tasks by automatically searching through thousands of documents and compiling the relevant information. “That gives lawyers more time to concentrate on the strategic side of their work,” says Grnarova. Most of today’s legal software focuses on one specific job without exploiting the parallels and synergies that exist between different tasks. In contrast, DeepJudge operates as a kind of all-round virtual assistant that supports lawyers with a wide range of tasks.
And that’s not the only thing that puts DeepJudge in a class of its own. Trained using millions of publicly available legal documents, the DeepJudge AI understands the semantic content of source documents and processes this in a context-sensitive way. Among other things, this enables the software to compare different contracts and automatically produce a template for a new contract. The platform can also “smartify” existing legal documents automatically by augmenting them with paragraphs from relevant laws and court rulings and adding references to existing registries of commerce. And it allows users to redact sensitive information with just one click.
« The artificial intelligence revolution was in full swing and it was becoming clear that computers could perform certain tasks to a superhuman degree » Paulina Grnarova
Understanding language add context
By drawing on a combination of deep learning and natural language processing, DeepJudge has made a quantum leap in semantic understanding and context awareness. “In 2017, I joined the Google Assistant research team at Google AI Language and stayed for two-and-a-half years. The experience I gained there has been invaluable,” says Grnarova. While working on the text-to-speech version of Google Assistant, she taught the artificial intelligence to condense a lengthy text and summarise the key points in just a few sentences. This required the AI to understand the content on the semantic level. Grnarova’s work on DeepJudge was also inspired by her internship in core machine learning at Google Brain, where she gained important insights into how models learn and what they look at to make decisions.
“I was incredibly fortunate to be working on my doctorate just as researchers were starting to develop deep learning and neural networks. The artificial intelligence revolution was in full swing and it was becoming clear that computers could perform certain tasks to a superhuman degree,” she says. DeepJudge is one of the first generation of ETH spin-offs to be supported by the AI Center. Grnarova’s inspiration for bringing AI to the legal profession came from her former ETH professor Thomas Hofmann, who had previously launched a successful start-up in this field.
New role as CEO
Prior to becoming CEO at DeepJudge, Grnarova had always helped write the code for every project and had primarily been in charge of the technical aspects. Now, however, she has an entirely new set of responsibilities, working at the interface between clients, investors and her employees. Her pitches and presentations no longer focus on technical issues; instead, they highlight the problems her software can solve and explain why customers should buy it. Grnarova has faced a steep learning curve as CEO – and her rapid success speaks volumes for her adaptability.
“I do miss programming sometimes, though the truth is that it gets a bit lonely writing your own code year after year, as I did for my doctoral thesis. That’s why it’s such a nice change to be working with a team as CEO,” she says. One year after founding DeepJudge, Grnarova already employs 14 people; during that time, she and her team have won awards and prizes worth over half a million Swiss francs. The philanthropic initiative Venture Kick has just injected another 150,000 Swiss francs into the company, and Grnarova recently gained a place on the prestigious Forbes 30 Under 30 list.
The Forbes accolade triggered a media storm that even reached her home country of North Macedonia, where her story was featured on national TV. Asked what she misses most about her country of birth, she can’t help smiling. “Family, friends and the social life that is such a big part of Macedonian culture – much more than it is here,” she says. She clearly believes in lifelong friendships built on solid foundations, and she thinks entrepreneurs need to be equally discerning in choosing the best people to collaborate with. “When you set up a company together, you inevitably have bad times as well as good, so it’s crucial to have a team of founders who support each other and have the same goals. That’s the secret to success; but, even more importantly, it’s what defines the kind of journey you share along the way,” she says before disappearing into the WOW Museum’s maze of mirrors.
This text appeared in the 22/01 issue of the ETH magazine Globe .