Bridging The Gap Between AI Research And Real-World Security - An interview With Srikanta Datta Tumkur

For Srikanta Datta Tumkur, the path to becoming a leader in AI-powered cybersecurity wasn't mapped out from the beginning, it emerged from years of building systems that couldn't afford to fail.

Growing up with a deep curiosity for technology and how systems work under pressure, Datta pursued advanced degrees at Carnegie Mellon University and HEC Paris, sharpening both his technical depth and business acumen along the way. But what truly defined his approach wasn't the classrooms - it was the trenches of building enterprise platforms where security wasn't optional, and where mistakes had real consequences.

"Security isn't something you add at the end," Datta reflects. "It has to be woven into the foundation. I learned that lesson early, building systems where a single vulnerability could mean catastrophe for thousands of users."

His career journey reads like a tour through the evolution of modern infrastructure. At Oracle, he helped build cloud platforms where compliance and hardening were non-negotiable. At Symphony, he designed a secure collaboration system that kept encryption keys entirely in customers' hands - a bold architectural choice that reflected his belief in giving users control. At Cisco, he contributed to enterprise network management at massive scale. Today, at Coupang, he leads GPU cloud infrastructure while championing security initiatives that have measurably reduced vulnerabilities across their systems.

Through each chapter, one theme persisted: the gap between what researchers write about and what actually works in production.

When Theory Meets the Real World

"I've spent years watching brilliant research ideas fail when they meet messy, real-world data," Datta explains. "Systems are noisy. Infrastructure keeps changing. Attackers don't follow the patterns you expect. Research has to survive those constraints to matter."

This philosophy drove Datta to pursue his own research, publishing peer-reviewed work on applying artificial intelligence to detect sophisticated cyber threats. His approach focuses on understanding how attackers move through systems - not through single obvious actions, but through subtle patterns that only emerge when you look at relationships between events, users, and systems.

"Traditional security tools look for known signatures," he says. "But the most dangerous threats don't announce themselves. They blend in. You have to teach systems to recognize suspicious patterns, not just known attacks."

A Breakthrough in Cyber-Physical Security

One of Datta's most significant recent contributions came in November 2025, when he presented research at the IEEE 12th International Conference on Cyber Security and Cloud Computing in New York City. The paper, titled "Explainable Multi-Modal Deep Learning Framework for Enhancing Cyber-Physical Systems Security," addressed a growing challenge that keeps security professionals awake at night: protecting the systems where digital technology meets the physical world.

Think of smart factories, power grids, medical devices, or autonomous vehicles. These cyber-physical systems blend software with physical processes, and a security breach doesn't just mean stolen data - it can mean real-world harm.

"The problem with most AI security tools is the black box issue," Datta explains. "They might detect a threat, but they can't tell you why. When you're protecting critical infrastructure, operators need to understand and trust the system's decisions."

Working with research collaborators, Datta developed a framework that combines multiple types of data—network traffic, sensor readings, system logs, and behavioral patterns - to identify threats while also explaining its reasoning in terms that human operators can verify and act upon.

"Explainability isn't a nice-to-have," Datta emphasizes. "In regulated industries, you can't deploy a system that says 'trust me.' You need to show your work."

This wasn't Datta's only contribution to the field in 2025. At the IEEE 5th Intelligent Cybersecurity Conference (ICSC), he presented "Neuro-Symbolic Approaches for Cybersecurity Policy Enforcement." The research tackled another persistent challenge: how to translate written security policies into systems that can actually enforce them.

"Security policies are often written in natural language," Datta explains. "But enforcement requires code. There's always a gap between what organizations say they do and what their systems actually do. We developed a hybrid approach that bridges that gap."

Looking Toward the Future

For someone with over fifteen filed patents to his name and recognition as an IEEE Senior Member, Datta remains remarkably focused on practical impact over accolades. He sees the next frontier of cybersecurity as one where AI helps defenders move faster than attackers, where security policies check themselves automatically, and where systems can recognize threats they've never seen before.

"The next major cybersecurity failure won't happen because we lacked tools," Datta warns. "It will happen because the surrounding infrastructure and governance weren't built with security in mind. The organizations that win will be the ones that treat security as a platform capability - not an afterthought."

In a field often divided between theorists and practitioners, Datta stands as a bridge builder - someone who believes the best security research is the kind that can survive contact with real systems, real data, and real attackers. His journey to presenting at IEEE conferences reflects a career built on one simple principle: ideas only matter when they work.

"Anyone can write a paper about a perfect solution," he says with a smile. "The real question is whether it still works when the data is messy, the attackers are creative, and your budget is tight. That's where the interesting problems are."

For organizations navigating the AI-powered future of cybersecurity, Datta's message is clear: the winners won't be those with the fanciest algorithms or the biggest budgets. They'll be the ones who understand that security is a foundation, not a feature - and who invest in people capable of translating cutting-edge research into systems that actually protect.

About Srikanta Datta Tumkur

Srikanta Datta Tumkur is a Director of AI Infrastructure with experience building enterprise platforms at Oracle, Symphony, Cisco, and Coupang. He holds degrees from Carnegie Mellon University and HEC Paris, and is an IEEE Senior Member.