I adapted this technique to work on title insurance risk, optimizing our machine learning algorithm to produce maximal value while enabling the best customer experience. It turns out, if you’re just optimizing a number, it doesn’t matter if it’s galactic dark matter or dollars of revenue and loss.
SAN FRANCISCO (PRWEB)
January 07, 2020
States Title Holding, Inc., one of the top ten largest real estate settlement service companies in the United States, was issued a patent for “Predictive Machine Learning Models” by the U.S. Patent and Trademark Office on December 17, 2019. The company can now find the “sweet spot” for automated title processing in its first-of-its-kind title insurance algorithm.
The patent for invention number 10,510,009 allows States Title to optimize the value-maximizing point in the instant underwriting model by modifying the previously well-known Markov Chain Monte Carlo technique with unique title-specific mathematical components. The company’s Chief Data Scientist, Andy Mahdavi, a Fulbright Scholar who earned his PhD in Astronomy and Astrophysics from Harvard University, previously applied the Markov technique to study dark matter using galaxy cluster data from NASA satellites.
“During the first half of my career, I was an astrophysicist, working on unraveling the nature of dark matter, a substance that makes up 85 percent of the cosmos and yet is totally invisible and extremely difficult, if not impossible, to detect in a lab. To do this, I leveraged a statistical technique called Markov Chain Monte Carlo, which optimizes theoretical models until they match the data, providing the most suitable match. When I arrived at States Title, I adapted this technique to work on title insurance risk, optimizing our machine learning algorithm to produce maximal value while enabling the best customer experience. It turns out, if you’re just optimizing a number, it doesn’t matter if it’s galactic dark matter or dollars of revenue and loss.”
– Andy Mahdavi, Chief Data Science Officer, States Title Holding, Inc.
This new patent complements the patent awarded to States Title in April 2019, which protects the company’s use of data science to create predictive title insurance by assigning a risk score to indicate how safe a property is in terms of potential liens or other liabilities that may cloud a property’s title. The second patent protects States Title’s unique application of the Markov technique to turn this risk score into a profit/loss number.
Mahdavi joined the company from Capital One, where he applied the scientific rigor of almost two decades in academia to develop hyper real-time fraud protection. He is emblematic of how States Title attracts experts from a variety of fields to disrupt an industry that has looked the same for more than 100 years.
“Anyone who has bought or refinanced a property has experienced the friction and frustration inherent to the traditional title and escrow process. Our team brings academic and industrial expertise from astrophysics, data science, machine intelligence, supply chain management, operations research and engineering disciplines together in a concerted effort to reimagine the residential mortgage industry.”
– Max Simkoff, Chief Executive Officer, States Title Holding, Inc.
About States Title Holding, Inc. and States Title
Founded in 2016 in San Francisco, States Title Holding, Inc. has grown rapidly to become one of the top ten largest title and escrow companies in the US. The States Title family of companies – States Title, North American Title Company (NATC) and North American Title Insurance Company (NATIC) – is able to impact 90% of all real estate transactions in the US. States Title’s first-of-its-kind, patented technology solution utilizes machine intelligence to make residential real estate closings vastly more simple and efficient. Current customers are able to close more loans, faster, and at a lower cost, while maintaining best in class quality and service. To learn more visit http://www.statestitle.com.
Share article on social media or email:
Published at Tue, 07 Jan 2020 00:00:00 +0000