Booz Allen spinout Modzy aims to bridge the ‘AI valley of death’

Artificial intelligence has been heralded as a game-changing technology for myriad sectors and applications. But deploying it effectively has proven to be continuously vexing.

According to Modzy CEO Josh Sullivan, a whopping 90% of machine learning models never even make it out of the lab into production, instead floundering in an “AI valley of death.” In his view, most commercial AI applications lack sufficient security, governance and other basics when first introduced.

Modzy is a platform for AI model operations and machine learning operations — aka ModelOps and MLOps — that launched within IT consulting company Booz Allen Hamilton in 2019, and was spun out into an independent company in December 2021.  Its goal is to solve implementation difficulties with AI, and help companies roll out AI applications efficiently and at scale, Sullivan told Technical.ly. The Vienna, Virginia-based company counts 17 employees and works with 25 partner companies, including Nvidia, Microsoft, Amazon and Snowflake.

Booz Allen holds a minority state in Modzy, which is part of a larger strategy to scale its technical work.

“Spinouts represent a maturation of our solutions business and demonstrates our commitment to invest in innovation at scale through identifying, evaluating, incubating and scaling businesses at the intersection of mission and technology,” Booz Allen EVP Frank DiGiammarino told Washington Technology. “It’s really our leading tech talent that develops these cutting-edge solutions like Modzy, and when they grow beyond experimentation, we identify the best opportunities to scale them further either inside or outside the firm, so they can reach their full potential.”

Modzy has three AI deployment modes, Sullivan said: Customers can run models on premise, in the cloud or “at the edge.” (The edge offering is now in private beta.) The platform is designed for revenue generation and cost reduction, as well as “better integration of AI applications, effectively bridging the AI valley of death, and accelerating models that have already been trained, as well as provide monitoring, governance, and security.”

The platform can explain just why an AI model made a certain prediction, and has drawn strong interest from the manufacturing sector, according to Sullivan, as it has been particularly effective for supply chain applications. An industrial vehicle manufacturer client, for instance, is using Modzy on the factory floor to predict factors affecting its supply chain, allowing employees to respond faster to problems that might arise.

Where is the technology being applied? Sullivan said Modzy has also been particularly well received in highly regulated markets, such as energy and telecommunications, where it’s used to enhance resiliency and assure security — say, 5G infrastructure companies. It’s also drawn interest from pharmaceutical and insurance companies, and a number of agencies across the Department of Defense, according to the CEO.

With a reported ability to speed the AI R&D process by 10%, Modzy is keeping a trained eye on its goal to bridge that valley of death.