In a time when AI is everywhere, data privacy has gone from a side concern to a top priority.
That’s the focus of CipherSonic Labs, a Boston-based startup co-founded by CISE Faculty Affiliate Professor Ajay Joshi (ECE) and BU alum Rashmi Agrawal (PhD ’23, ECE) in January 2024..
As companies across verticals (healthcare, financial services, supply chain, education, etc.) race to adopt AI, many are facing the same issue—how to protect sensitive information like training data, user inputs, and even the AI models themselves.
For example, in March, Yale New Haven Health detected unauthorized access to its network, leading to a leak of data belonging to approximately 5.6 million patients. The compromised information varied by individual but included names, dates of birth, contact details, Social Security numbers, patient types, and medical record numbers. However, the electronic medical record system and financial data remained unaffected.
Yale New Haven Health contained the incident, initiated an investigation with cybersecurity experts, and notified law enforcement, but these are the types of incidents that CiperSonic Labs is working to prevent.
“There are data breaches happening all the time,” Joshi said. “The consequences, like loss of business, credibility, trust, can be devastating.”
CipherSonic’s story starts with a combination of need and timing.
“That need came more from the world of AI,” Joshi explained. “These companies face challenges protecting model, training, and user data.”
At the same time, Rashmi Agrawal was doing research at BU on encrypted data processing—specifically, how to keep data encrypted while still being able to use it.
Encryption itself isn’t new, but making it practical across storage, transfer, and computation has always been difficult. Agrawal solved that problem, and CipherSonic was born.
“One of the key breakthroughs I worked on during my Ph.D. was solving the performance bottleneck of Fully Homomorphic Encryption, a technology that allows computation directly on encrypted data,” Agrawal, a 2023 EECS Rising Star, said.
“We achieved over a thousand times speedup, making FHE viable for real-world use. This innovation forms the core of what we are building at CipherSonic: enabling AI models to run on encrypted data while protecting both user privacy and model IP,” she said.
“We didn’t want this to just sit in a journal,” Joshi said. “We wanted to commercialize it so the world could benefit from it.”
Over the years, CipherSonic has won or been a finalist for numerous prestigious awards, including the BU Ignition Award, MassChallenge RESOLVE ’25, an accelerator for startups with a focus on social impact.
That journey from lab to startup hasn’t been easy. One of the biggest surprises? The gap between research and real-world use. “Doing research and building a product are completely different worlds,” Joshi said. “You can publish groundbreaking work, but getting a customer to pay for it takes a different skill set entirely.”
CipherSonic is a B2B company that sells to other businesses rather than individual consumers. That makes the sales process longer and more complicated.
“We have to go talk to other businesses, convince them how our product will help them, and how they should use it to future-proof their company,” Joshi explained.
Even so, the long-term vision goes beyond selling software. CipherSonic is trying to unlock a safer, faster way for organizations to collaborate with sensitive data. For example, Joshi described a situation where a hospital and a research lab could share encrypted patient data without compromising privacy.
“[Hospitals] have patient data, and say there is a research lab, separate from the hospital, which is developing new drugs. So they want to share data, but it is challenging due to HIPAA regulations, privacy regulations,” Joshi said. “With our technology, the patient data is encrypted. Everything is happening in the encrypted world. Patient privacy is protected.”
CipherSonic is still early in its journey, but growth is already underway. They’re aiming to double their team size to six by end of summer.
“The plan is to use that momentum to sign up more customers and raise more money,” Joshi said.
Ultimately, Joshi believes CipherSonic’s work is about balancing privacy with progress. “Finding a way to provide value and good service to the customer without compromising their privacy,” he said. “It will be super important because we live in a data-driven world.”
“All these benefits are basically unlocked because of this new technology, which can lead to better customer service, more value, and better business intelligence,” he added.
CipherSonic isn’t just protecting data—it’s also pushing the conversation forward about how we use it responsibly in the age of AI.
“I’m especially excited about how our platform helps industries unlock AI’s full potential while staying secure and compliant. It’s been incredibly rewarding to translate deep research into a product that can truly shift how enterprises think about data protection,” Agrawal said.”
—
Ajay Joshi is a professor in Boston University’s Electrical and Computer Engineering (ECE) department, where he heads the Integrated Circuits, Architectures, and Systems Group. His work spans computer architecture, hardware security, VLSI design, and silicon photonics.
Before coming to BU, he earned his Ph.D. from Georgia Tech in 2006 and completed postdoctoral research at MIT. He also spent time as a Visiting Research Scientist at Google and an Architect at Lightmatter.
Professor Joshi has picked up several honors over the years, including the NSF CAREER Award, BU ECE’s Award for Excellence in Teaching, the BU Ignition Award, a Best Paper Award at ASIACCS 2018 and HOST 2023, and Google Faculty Awards in both 2018 and 2019. Right now, he’s also serving as an Associate Editor for the IEEE Transactions on VLSI Systems.