Deniz Johnson
Chief Operating Officer, Stratyfy
Years experience:
40 years

Deniz Johnson shifted from healthcare technology to financial tech 24 years ago, and has been solving complex problems in the highly regulated industry ever since. She currently serves as the COO of Stratyfy, an artificial intelligence company with a proprietary AI technology that Johnson says removes hidden bias in data and models. She also founded Pera-Partners, a management consulting company specializing in fintech adoption and digital transformation. Born and raised in Turkey, Johnson is known as a thought leader in the U.S. fintech landscape, and is passionate about the ways responsible uses of AI and machine learning in highly regulated industries can increase inclusivity in the business world.

Stratyfy was recently selected as the technology provider for the Underwriting for Racial Justice initiative, which brought 21 banks, credit unions and CDFIs – including Boston-based Berkshire Bank and Eastern Bank – together to explore and pilot ways to make loan underwriting less racially biased.

Q: What are the ways Stratyfy’s technology can help banks and credit unions to improve lending to people of color?
Our mission is to help financial institutions make better decisions through providing technology that enables financial inclusion. Stratyfy is an AI company and our focus is to help these institutions improve their lending practices through looking at data in order to enrich communities that might have been traditionally underrepresented.

We have three technology products: one is PRE, a patent-pending technology for loan origination that is interpretable and transparent, and allows you to understand and interact with data. We also tap the expertise of the humans along with the strength of artificial intelligence and machine learning to make predictions that are unique to the institutions, their data and goals. Another is called UnBias and it looks at machine learning models and determines and uncovers bias, provides insights to understand [data] and gives capabilities to undo that bias. The last product looks at fraud detection, and we work with fraud detection companies specifically looking at fraudulent activities on the transaction side.

The technology allows [Underwriting for Racial Justice participants] to see and understand their lending data, at the same time interact with their own data, and have all the dials with the models we create for them. We are also measuring the impact of the data and the learnings are being shared among all of the participants.

Q: What are the intricacies of Stratyfy’s technology? Will the technology require a different “plumbing” to get it into banks’ systems? 
Our technology is transparent and interpretable. It means the machine is learning what the human is doing, making a decision like the human, and the human can at any time see what is triggering and impacting those decisions. This is now a very powerful tool for a financial institution to drive scalability and efficiency in understanding their data and actually meeting their goals.

The banking industry has some legacy products that you have to keep, which means it’s hard to integrate and do interfacing. You have all these “little mushrooms” or different products that kind of go around your legacy products. Or if you’re not that size of a company, you’ll have core and then everything else is smaller components. If you can build an API or if you have an API connection, that’s all you need to connect to us. We can be the bank’s core decisioning engine or we can supplement their decisioning engine and be the decision optimization. We can be a part of the ecosystem of a bank as we are a cloud-based product.

This can then integrate into our loan origination system, and the standard APIs that we work with, we use banks’ data. We never own their data, which can be loan application data and other supplemental information from credit bureaus and other data partners that can influence the decision to originate and underwrite loans. For URJ, we look at loan origination data which banks use to make a decision on a loan.

The beauty of our loan origination product is the human touch, where it is part of the process that we have human to human conversations with the loan officers, asking them what are they looking at when making a decision [to do lending]. That’s the part of our expertise that is not captured in the data. We capture these conversations and understand their data. And that is being fed into the AI and machine learning technology, and banks can see their data and understand them.

Q: How can Stratyfy be unbiased when we know that so much of our lives are deeply influenced by our racial and class backgrounds?
For the financial institutions under URJ that committed to make modifications into their lending practices to reach minority communities, each lender has a focus and goal. We look at all their lending data, and understanding how they’re making lending decisions. Each data is tagged anonymously and it shared with the group as a “consortium data.” We can tell through the data and the human expertise how the lenders make decisions and what influences their decisions. Stratyfy and all of the other lenders are trying to work together, looking at different cases, and understanding their decision making and learning from each other.

Data science is usually looking at data from past decades, analyzing that data, and basing decisions for the future from past data. But people like me, who I like to call a “transplant” and not an immigrant, are not factored in. Women and people of color are not considered in historical data. For me, it is really important to make use of the technology we have today to see what is currently happening and make decisions from the present.

Using our UnBias product, we analyzed the U.S. home mortgage data in 2021 and found that in roughly 1 million conventional home mortgage loans, 1,117 Black applicants were unjustly denied mortgages. This translates into $387 million in credit denials to Black borrowers due to unjust denials. In Massachusetts, we have a history of redlining, which was acceptable then but not acceptable now. With our technology today, we can measure how much impact we’re making in the [affected] communities because we can see past and future decisions and using this to make strides in impacting communities in a positive way.

Johnson’s Five Favorite Books

  1. “Weapons of Math Destruction,” by Cathy O’Neil
  2. “The No Asshole Rule,” by Robert Sutton
  3. “The Archer,” by Paulo Coelho
  4. “My Sweet Orange Tree,” by Jose Mauro De Vasconcelos
  5. “Reimagining Capitalism in a World on Fire,” by Rebecca Henderson

Solving Lending Disparities One Data Point at a Time

by Nika Cataldo time to read: 4 min