Akhilesh Raj
Ph.D. Candidate · Reinforcement Learning · HPC · Cybersecurity
I build AI-driven systems that make computers smarter about their own power consumption and make critical infrastructure more resilient to cyber threats. My work sits at the intersection of offline reinforcement learning, high-performance computing, and cyber-physical security — bridging theory and production-scale deployment at national labs.
Focus Areas
Research
Applying offline reinforcement learning to design autonomous CPU power controllers using Intel RAPL and hardware performance counters. Achieves measurable energy savings with minimal degradation across compute- and memory-bound workloads.
Building advanced testbeds to evaluate attack-resilient RL agents in industrial OT environments. Uses OpenStack-based network emulation to simulate real-world SCADA scenarios and demonstrate detection and mitigation of adversarial attacks.
Developing performance-aware energy management for 5G base stations with RL, and AI@Edge-to-HPC infrastructure (Charon) that connects distributed intelligent workloads to large-scale HPC resources seamlessly.
Highlighted Work
Featured Projects
Application-agnostic offline RL agent using Intel RAPL hardware counters and heartbeat signals to cut HPC node energy use by ~23% with negligible performance loss.
OpenStack-based industrial network emulator for training and evaluating blue-team RL agents defending against adversarial attacks on SCADA and ICS infrastructure.
Prototyped an IoT-integrated smart vault with secure remote access control, real-time delivery notifications, and tamper detection for smart-home environments.
Tools & Technologies
Tech Stack
Latest Updates
News & Milestones
- 2025 Paper accepted — "Application-Architecture Agnostic Efficiency Optimization using Reinforcement Learning" — architecture-agnostic RL approach for HPC energy management. [update with venue/link]
- 2024 Collaboration continued with Argonne National Laboratory on energy-efficient HPC architectures and advanced cybersecurity testbeds for OT networks.
- 2024 Ph.D. candidacy achieved at Vanderbilt University, Department of Electrical and Computer Engineering. Research advisory: offline RL meets HPC systems.
- 2023 New publication — "Reinforcement Learning-based Performance-aware Energy Management in 5G Base Stations" — demonstrating RL-driven 5G energy efficiency gains.
- 2023 Charon paper — "An End-to-End Infrastructure for Connecting AI@Edge to HPC" presented, bridging edge AI deployments to large-scale HPC resources.
- 2022 Testbed demo — Autonomous cybersecurity testbed for OT networks showcased. Live demos include a chemical plant and power grid scenario under adversarial conditions. [update with venue]
Live Demonstrations
Project Demos
Affiliations & Partners
Collaborations
Let's Connect
Contact
I'm always happy to discuss research collaborations, new ideas at the intersection of AI and systems, or PhD/postdoc opportunities. The best way to reach me is by email.