Sep 5, 2024

The startup is developing industry-specific models for pharma, materials, and other industries. PatSnap, a Singaporean startup that helps businesses search for and analyze patent information, is starting to offer more services tailored to specific industries by increasingly tapping generative artificial intelligence. Co-founder Guan Dian told Nikkei Asia in a recent interview that the company—whose more than 12,000 corporate clients include Tesla, Siemens, and NASA—aims to win more customers across fields such as pharmaceuticals and material sciences through the new services. Patsnap’s generative AI uses chatbots to answer questions from researchers and IP lawyers, helping them understand complex patent documentation in different regions—something that could otherwise be a huge undertaking when companies are looking to launch products or enter new markets. The services also help clients track any recent technological advances by their competitors. The startup’s large language model (LLM) has been trained using Meta’s open-source AI model after accumulating data from 200 million patents, a million scientific and technical books, and billions of new articles and other sources. “It doesn’t make sense for a smaller company like us to build another foundation level [of AI] because it’s very costly,” Guan said, referring to dominant generative AI technology developed by US companies like OpenAI, Google, and Meta. She described PatSnap’s AI model as a “specialist” that can help experts in different fields seek more technical information, whereas she said existing services like ChatGPT are more for wider general knowledge. “It’s time for companies with expertise in particular verticals—whether it’s in patents, medicine, or marketing—to start building on top of these language models,” she said. Since embarking on its own generative AI development a year ago, the company this year launched services for general IPs and another targeted at patents in the pharmaceutical industry, Guan said, adding that a similar offering for the material sciences sector is in the pipeline. Patsnap was founded by CEO Jeffrey Tiong in 2007 after graduating from the National University of Singapore. Tiong, a biomedical engineer, got the product idea after discovering the difficulty of analyzing patents while interning at a medical startup in the US. Guan, who studied computer science and joined shortly afterward as a co-founder, now leads its Asian businesses. Early investors include a fund under  Vertex Holdings , the venture capital arm of Singapore’s state investor Temasek Holdings. In 2021, the company became a unicorn—an unlisted company valued at USD 1 billion or more—after a USD 300 million fundraising led by SoftBank Group’s Vision Fund and China’s Tencent Holdings. Prominent venture capital firm Sequoia Capital of the US has also invested in the company. Guan said the company has reached an annual recurring revenue—a key metric for subscription-based software companies—of USD 100 million and turned profitable earlier this year. Its next focus, Guan said, includes getting ready for an initial public offering. “We still have, in our mind, at least two to three years before the actual IPO. But again, everything is about market timing, and things might change.” According to Guan, US and Chinese clients currently account for around 70% of the company’s revenue. She added that PatSnap is expanding its team in Japan, hoping to boost the country’s contribution to its revenues to a share of about 10% from just a fraction at the moment. Guan said the company is benefiting from “decoupling” in the global supply chain, especially the competition between American and Chinese firms, as it is “forcing each side to invest more” in R&D. “That means more R&D expenditure and higher demand for our kinds of products.” This article first appeared on  Nikkei Asia . It has been republished here as part of 36Kr’s ongoing  partnership with Nikkei . Share