Desi wins $25 million for technology that makes AI models more efficient

- Advertisement -

Desi, a 50-employee start-up company that develops a platform for building and optimizing AI systems, today announced the closing of a $25 million Series B funding round led by Insight Partners with participation from Square Peg, Emerge, Jibe Ventures, Fort Ross. Ventures and ICON, bringing the company’s total to $55.1 million. The funds will be used to expand DesiAccording to co-founder and CEO Jonathan Geifman, the company’s activity is to enter the market, as well as support the company’s research and development.

- Advertisement -

Companies face a number of barriers when building AI models for text, audio and image analysis to deploy in their applications and services. The cost is too high – training one model on commercial equipment can cost tens of thousands dollars, if not more. While new generations of chips and custom-designed AI accelerators have helped ease the load somewhat, building a model from scratch is still a challenge.

- Advertisement -

Geifman suggests Neural Architecture Search (NAS) as a solution. NAS, a family of methods that Desi relies heavily on, can help automatically find low-cost optimal models for a given problem. Desi is not unique in this – Google Apex AI service uses NAS to optimize the performance of models for specific customer-specified tasks. But Geifman says the Deci platform offers access to NAS capabilities at a lower cost.

In 2019 Geifman co-founded Deci with Ran El Yaniv and entrepreneur Jonathan Elial. Geifman and El-Yaniv met at the Technion’s computer science department, where Geifman was a Ph.D. and El-Yaniv was a professor.

- Advertisement -

“Desipatented technology [can generate] new image classification models that…provide more than a twofold improvement in runtime combined with improved accuracy over the most powerful publicly available models,” Geifman told TechCrunch via email. “This means that AI applications that previously could only be deployed on large and expensive GPUs can now be deployed on CPUs.”


Image credits: Desi

These are high demands. But Deci has the backing of Intel, which last March announced a strategic business and technology collaboration with a startup to optimize machine learning on Intel processors. The partnership resulted in a model that speeds up question-answering tasks on Intel processors and an image classification model for Cascade Lake processors that “significantly reduces computational overhead,” says Geifman.

Geifman previously told TechCrunch that one of Deci’s clients, a videoconferencing provider, used the platform to roll out a feature that blurs backgrounds on users’ devices. Others have used Deci to build better models for their own internal computing, even if they theoretically have the GPUs and processing power to do just about anything.

Desi was created to empower developers and eliminate manufacturing bottlenecks throughout the AI ​​lifecycle,” Geifman said. “The business impact of this opportunity is… reduced production time and the ability to discover new use cases for AI and reach new market segments on resource-constrained devices.”

Geifman also notes that compressed models can help companies save on inference computation costs, the cost of actually maintaining the models once they’ve been deployed. Due in part to the popularity of cloud hosting models, more than a third of companies regularly experience cloud budget overruns of up to 40%. interview surveillance software provider Pepperdata.

While Geifman says Deci’s business continues to grow, the startup is facing challenges, including the technical limitations of NAS. (NAS, which difficult to assesscan be costly and time consuming.) In addition, Deci also competes with a number of companies developing ways to make models more efficient, such as OctoML, Neural Magic, and OmniML.

The coming months will be a test of Deci’s resistance to headwinds.

“While we cannot disclose the estimate, we can say that it has increased significantly from the previous round. Due to the growth of Deci’s business and the opportunity to expand the product into additional areas such as natural language processing, our existing investors have decided to redouble their efforts to support this growth,” Geifman said. “We have not seen a significant impact [from recent economic developments]. Our focus was mainly on enterprises, while the slowdown mainly affected mid-sized companies and startups.”

Insight Partners Managing Director Lonn Jaffe, Deci Board Member, added in an email to TechCrunch: “Deci’s powerful technology allows you to bring in your AI models, data, and target hardware—whether that hardware is on the edge or in the cloud—and will help you to find alternative models that will generate similar prediction accuracy with significantly higher efficiency… [It’s a value add because] hourCreating a more efficient infrastructure for AI systems can make AI products qualitatively different and better, not just cheaper and faster to operate.”

Credit: /

- Advertisement -

Stay on top - Get the daily news in your inbox

DMCA / Correction Notice

Recent Articles

Related Stories

Stay on top - Get the daily news in your inbox