Meet Gemsen — A Demo Day 12 Presenting Startup

This year, Fintech Sandbox Demo Day will take place on Tuesday, April 28. The presentations will be virtual and the event, as always, is free. Demo Days are exciting because we get to showcase startups that are on the very cutting edge of innovation and you get to see what they’re up to before they’re discovered.

Over the next few weeks, we’ll highlight this year’s presenting entrepreneurs. Today, we’re talking to Charlie Ko, Co-founder and CEO of Boston-based fintech Gemsen. Gemsen is capable of processing massive amounts of data efficiently and securely, producing output that is accurate and in a usable form. Gemsen tames the data deluge using prediction models that require less compute and storage than other technologies.

Gemsen allows organizations to harness the power of machine learning without exposing sensitive data, ensuring complete compliance with data protection regulations. Whether applied within workflows such as insurtech, regtech, fraud detection, credit analytics, compliance, or investment research, Gemsen enables secure collaboration throughout the entire process.

Charlie, tell us a bit about Gemsen. What problems are you solving? What makes your approach different?

Most of today’s advanced technologies are extremely resource hungry. To perform at the speed of current data streams, they often require massive compute and storage in the form of GPU server farms. We took a completely different approach to use math to simplify the computations to what is essential. This not only preserves predictive performance, but also dramatically reduces the required resources by 99% or greater.

What is your company’s origin story?

We are driven by simplicity and elegance. For nearly two decades, our founding team has tackled complex and large-scale data problems by distilling information from noise. We have always understood that data is only useful if patterns can be detected quickly and reliably.

What does it mean to say that you capture the essential statistical structure of data in motion?

When deconstructing the AI/ML models, the functions of these ‘machines’ are distilled into essential equations. Even the most sophisticated models are defined by equations that perform the work of finding patterns. Many models can be simplified. Some models can be captured and interpreted through pre-computations that reflect the inter-relationships of the input data. These relationships can be described via higher and lower order statistical moments. The key is finding the sufficient set of statistics to produce a lossless rendition of the data for the relevant ML models in question.

This essential structure is the key to extremely fast ML training and inference. Additional important outcomes of this design are Gemsen’s adaptability to streaming data, in addition to interpretability and transparency of the models themselves.

Who is Gemsen’s target market?

We target financial institutions that deal with large-scale streaming data volumes each day. Sample institutions include banks, investment firms, brokers, payment processors, and insurance companies. Sample workflows include fraud detection, customer prediction, securities trading, and credit analytics.

Which Fintech Sandbox Data Partners have you worked with?

massive.com provides an extremely rich data set of market flow information spanning many years. Since data is the fuel that powers AI/ML, their clean and reliable data made testing easier for Gemsen. Billions of rows of data flowing at millisecond speed helped us prove our reliability and improve our capabilities.

What milestones has Gemsen achieved so far?

Our software is already enterprise scale and has been deployed in Fortune 100 companies outside finance. There, we have demonstrated substantial increase in efficiency and reliability of the technology. Gemsen was founded to bring this technology to finance and is in active testing within the finance domain.

Can you describe what it’s been like to be part of the Fintech Sandbox community?

It’s been an amazing experience. The people in this community have been extremely giving of their time and expertise. The best part is everyone’s approachability. Coupled with their willingness to share their experience and ideas, this makes a powerful combination.

You also took part in the MassChallenge Fintech Program last year. What was that like?

It was great from the start. The number of conversations and connections from day 1 was extremely important for us to gain quick momentum. I can’t overstate how helpful the entire ecosystem has been for Gemsen. In fact, the connections brought us include two key advisors. Both helped gain resources that are critical in our ongoing work. This is on-top of getting an inside track to Fintech Week and other high-profile events throughout the year, which nets more ideas and connections.

What’s next for Gemsen?

Just spreading the word – that you can tame the data deluge while still being reliable, accurate and efficient.

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

To hear more about Gemsen and five additional exciting fintech startups, be sure to register for Fintech Sandbox Demo Day 12!