Sumit J. Darak
Papers
3
Total Citations
28
H-Index
3
About
Sumit J. Darak is a leading researcher in the design of intelligent, reconfigurable architectures for multi-armed bandit (MAB) algorithms, with a focus on enabling real-time decision-making on resource-constrained edge devices. His work bridges the gap between theoretical machine learning and practical hardware implementation, particularly for applications in wireless communications, the Internet of Things (IoT), and robotics. Darak’s major contributions include developing novel, hardware-efficient architectures that allow MAB algorithms—which optimally balance exploration and exploitation without prior training—to be deployed on system-on-chip platforms like Zynq. His 2020 paper on a reconfigurable architecture for KL divergence-based MAB algorithms (12 citations) and his 2022 comparative study of frequentist versus Bayesian approaches on Zynq SoCs (11 citations) are among his most cited works, demonstrating the impact of his research. By creating computationally efficient and reconfigurable designs, Darak has advanced the practical viability of MAB algorithms for autonomous systems, enabling them to adapt to unknown environments with minimal latency and power consumption. His work is essential for students and researchers interested in the intersection of reinforcement learning, embedded systems, and adaptive hardware.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3