Cognaize raises $18M to build a better LLM for the finance sector, one that keeps humans in the loop | TechCrunch
4 min read![](https://sildenafilxu.com/wp-content/uploads/2023/07/money-explosion-gradient-1024x576.jpg)
[ad_1]
Artificial intelligence has had a major impact over the years in the finance sector, helping neobanks personalize services for customers, lenders evaluate loan applications, digital providers detect fraud and security issues, analysts run predictive modelling for investments and more.
Yet much of the work done today is in the area of structured data. With a wave of unstructured data waiting to be tapped and used in the process, a New York startup called Cognaize is taking a hybrid approach. It’s built a platform for the processing of unstructured data for financial AI applications, and it complements that with “humans in the loop” to refine the work.
And today, it is announcing $18 million in funding to expand its business on the back of winning some big customers for its services. Customers include two of the three biggest credit ratings agencies, large insurance companies and financial services businesses.
Argonautic Ventures is leading the round with Metaplanet and other unnamed investors are also participating. Cognaize is not disclosing its valuation except to confirm that it is in range of hundreds of millions.
The funding will be going towards hiring — Cognaize has offices also in Germany and Armenia — research, product development and business development.
Cognaize’s founder Vahe Andonians is also the CTO and CPO, and he previously founded another fintech that provided analytics and risk management around credit investments that was eventually acquired by Moody’s. His approach there and with Cognaize is based around the idea that AI may be able to do things that humans cannot, but all the same it cannot replace humans.
The premise that Cognaize is taking is that while there is a seemingly limitless amount of data available to the finance industry these days to gain better insights about their services, the state of the market and their customers, it typically only uses a small proportion of that data, the structured part.
The startup has built a platform that taps deep learning trained specifically on financial models and a very wide variety of documents — 1.3 million in all — that might have many different “cells” of information on them requiring a more expert eye to “read.” (The documents cover loan applications, but also SEC filings, ESG documents, presentations, trustee reports and more.)
That platform, in turn, is used by human workers, typically financial analysts, to both help correct what is being read, and to make conclusions and decisions based not the outcomes.
“If you are a bank you have three options right now,” said Al Eisaian, Cognaize’s CEO. “You can try to build AI capabilities in-house but forget that. You can go through the general AI model for example using ChatGPT and try to implement it using an army of consultants. Or option three is us. We enable and educate you.”
Eisaian, a repeat enterprise founder with exits to companies like VMWare in his background, is not a founder of Cognaize but joined very soon after Cognaize got off the ground. The reason for the delay was because he needed to find a successor at the last company he founded and was leading, an aerial imagery analytics specialist called Intelinair.
The growth of startups like Cognaize in the field of AI highlights an important theme in the space: while there will likely be a number of companies like OpenAI, Google, Anthropic and others making big swings at general knowledge graphs in AI, building genuinely “large” large language models in the process, there is an equally interesting trend of very strong players focusing on specific fields and use cases. Those players may still be building “large” LLMs, but they are more focused on being deep than wide in their scope.
Yes, the biggest of them all may well try to do both, but specialists may always be able to speak the language of their customers are little more directly, and that might be what investors are betting on, too.
“We are thrilled to partner with Cognaize as they apply the transformative power of AI and large language models (LLMs) to finance,“ said Viken Douzdjian, managing Partner at argonautic Ventures, in a statement. “AI has disrupted various industries, but the massive amount of unstructured financial data creates countless use cases that need finance-specific, generative AI. The Cognaize platform can process vast amounts of unstructured financial data and extract insights with remarkable precision and speed, resulting in enhanced decision-making, risk assessment, and the uncovering of patterns and trends previously obscured by complexity and human error. We have strong conviction in Al, Vahe, and the Cognaize team to define how the finance industry interacts with AI.”
“Cognaize is a company to watch as they are one of the first to deliver repeatable and measurable value through AI in the financial industry. It was an easy decision to invest in Al, Vahe, and the entire Cognaize team,” added Rauno Miljand, managing partner, Metaplanet. “They have already harnessed the power of AI as demonstrated by the enviable growth of Cognaize’s business, the global leaders in finance that they have already secured as customers, and their unmatched technology roadmap. They are rapidly redefining how the finance industry can leverage modern AI to harness the power of their own data to dramatically cut costs while simultaneously creating new competitive advantage.”
The most convincing arguments for more targeted approaches, of course, are that they will give better results and be trained on specifically what a company needs; but also, they may turn out to be less expensive to run, given the smaller parameters of their LLMs requiring less compute power.
“There are always going to be opportunities because we are more agile and focused,” Andonians said. “That gives us the edge.”
“Having said that, only the paranoids survive and so we are leveraging things like ChatGPT too where it makes sense,” he added after a pause.
[ad_2]
Source link