Raghav Vadhera
Biography
Dr. Raghav Vadhera is a highly accomplished technology leader and AI/ML architect with over two decades of experience driving innovation across defense, healthcare, finance, higher education, and aerospace industries. He currently serves as a Principal AI Scientist at Raytheon Intelligence & Space, where he leads next-generation artificial intelligence, machine learning, and cyber analytics initiatives to enhance national defense capabilities and public safety infrastructure.
With dual PhDs in Cyber Analytics (The George Washington University) and Artificial Intelligence/Machine Learning (University of Texas), and an M.Tech from IIT Delhi, Dr. Vadhera combines deep technical acumen with strategic vision. His pioneering work includes the development of real-time AI weather models for NOAA?reducing model execution time from 3 hours to 5 minutes?and autonomous agentic drone systems capable of intelligent, context-aware mission execution.
Dr. Vadhera?s career spans leadership roles at institutions such as MIT Lincoln Laboratory, BAE Systems, Harvard University, and Fidelity Investments. His current research focuses on modular neural networks, reinforcement learning, and AI for predictive threat detection in cybersecurity. He is a thought leader recognized with numerous accolades, including the 2025 NTCS Innovation Award and the RI&S Innovators Award.
He is a keynote speaker at global AI/ML conferences, editorial board member for Discover Internet of Things (Springer Nature), and serves on the Industry Advisory Board for the University of Texas Department of Computer Science. He holds active U.S. government Top Secret/SCI clearance with polygraph.
Research Interest
Cyber Machine Learning & AI to Stop the most sophisticated cyberattacks using artificial intelligence & machine
learning. Design and implement threat detection systems by using ML to identify meaningful patterns in large disparate datasets by
understanding threat vectors and potential vulnerabilities.
Abstract
?Neural Network Architecture Optimization Using Reinforcement Learning For Cyber Resilience?
Static, non-adaptive cyber resilience algorithms fall short in countering today?s fast-evolving modern cyber attack threats. This gap highlights the urgent need for advanced, real-time adaptive systems and predictive solutions capable of learning and responding to both known and emerging cyber attacks. Recent advances in Artificial Intelligence (AI) and Network Optimizated Architecture ( NOA) offer significant potential to tackle complex cyber challenges. Among these, Reinforcement Learning (RL) has demonstrated its efficacy in incrementally building and optimizing neural network classifiers, enabling systems to dynamically adapt to new data and evolving threats (Vadhera_Huber_2023). By leveraging RL , this research seeks to reduce the dependency on frequent retraining, thereby facilitating a continuous learning-based cyber defense strategy. This research integrates RL into NOA to achieve automated tuning and introspective enhancement of neural networks. The proposed framework emphasizes continuous adaptation, enabling a resilient defense against evolving cyber threats in real time. In addition, a framework for learning cyber resilience policies and the embedding of abstract problems are introduced, allowing the transfer of learned policies, thus strengthening the system?s capacity to combat new and unexpected cyberattacks. This approach also incorporates the dynamic construction of heterogeneous network architectures to address the rapidly changing landscape of cyber threats and vulnerabilities. The growing availability of data on global cyber attacks, vulnerabilities, and emerging trends underscores the urgent need for solutions that can incrementally and effectively learn from this dynamic threat landscape?especially in the defense sector. Traditional algorithms often fall short in addressing such complexity. To bridge this gap, this research explores a reinforcement learning (RL)-based policy for developing optimized neural network architectures tailored to classification and regression tasks. Building on the research of Vadhera (Vadhera_Huber_2023) and applying it to defense-specific data?including cyber attacks and vulnerabilities?this work introduces a robust cyber-resilience framework. The framework generates optimized network architectures designed to defend against unprecedented threats. By continuously enhancing neural network models, this research lays the groundwork for a more resilient cyber defense posture, capable of withstanding both known and emerging challenges in an ever-evolving threat landscape.
