SRI spins out Kurve to help customers automate AI to speed up business operations.
Wes Madrigal is CEO and co-founder of Kurve, a Miami-based company working to transform the labor-intensive process of readying data for AI analysis. Here he talks about his time as an SRI entrepreneur in residence (EIR), his experience working with SRI, and the excitement of an unexpected journey.
—
I was born in Chicago and grew up in a small rural town near St. Louis, Missouri. I was always interested in technology and thought that I might want to go into actuarial science, which uses risk-based math models for the insurance industry. But in college, I kept changing my mind.
My older brother, a software engineer, thought Iād enjoy working in his field, and suggested I learn about coding software, so I taught myself and was hooked. In 2013, I started working as a data science intern at a startup founded by Rayid Ghani, who had been the chief data scientist for Barack Obamaās 2012 presidential campaign. Rayid had also started a program called Data Science for Social Good, which brought data scientists from around the world to Chicago to tackle problems in fields like education, public health, and public safety. Our co-located offices allowed me to see how data could help society, and how I could use my skills to have impact in the world. I was fascinated by the size and scope of the data we worked with, and how even a small amount of model intelligence over billions of data points had potential for positive change.
Early machine learning for building labor-intensive models
By 2016, people were talking about machine learning operations ā or MLOps ā and starting to focus on the gap between traditional software engineering and data science, which was the niche Iād been working in for several years. With increasing demand for these skills, I launched my first company, Mad Consulting, in 2020 to help businesses transition algorithms to production systems. The problem that my customers faced was that in order to extract value from data, they must be able to understand what the data is, what it contains, what it describes, and how it connects to the companyās application.
I was being asked by clients to build models to forecast revenue or predict when a customer would leave. That often meant going through hundreds or thousands of data sets to build just one model. Itās a laborious problem that canāt be solved by adding more people.
Connecting to SRIās experts
A friend had successfully launched a couple of companies with SRI. He introduced me to its ventures team, and I learned that they had encountered some of the same issues in their own workflows. The problem resonated with them, and we decided to tackle it by incubating Kurve together. I quickly became an EIR, and I was able to work with Ajay Divakaran, senior technical director of SRIās Vision and Learning lab, and Ryan Lewis, senior director of the Ventures team.
SRIās expert team developed new methods that leverage generative AI, specifically causal inference techniques, to automatically extract metadata from disparate, uncatalogued data sources. These generated data are essential for customers seeking to train and deploy enterprise AI tools, such as chatbots or enhanced search functionality, across proprietary data.
Itās not every day that you get to work with such a premier organization like SRI with its history and experience, experts and labs, and vast network. It has been phenomenal. Solving this problem can help organizations monetize AI, something that is on everyoneās mind. Roughly 40 percent of development time is spent on manually mapping and integrating to optimize AI for specific applications. If it can be automated, it would lower costs by freeing people to work on much higher-value tasks.
Kurveās mission is to take raw data and automate the process to make it usable for analysis and AI. Today, there are workers all over the world in a range of industries struggling to understand how to map their data to specific use cases, especially as the world starts to integrate AI. Imagine data as a terrain of information. What Kurve is trying to accomplish is to show an overlay of a map onto that terrain to help customers see what they have, how it connects, and how to navigate it to extract value.
āItās an extremely exciting time for the AI industry, and certainly for me and my team,ā said Madrigal.