Raj Adhikari - ECE PhD Student of the Month - November 2025
Raj Adhikari is a Ph.D. student in the Electrical and Computer Engineering department at NJIT, advised by Dr. Tao Han. His research focuses on enhancing the efficiency and reliability of data transmission and computation in next-generation intelligent systems such as autonomous vehicles and edge–cloud networks. His aim is to design methods that enable these systems to operate seamlessly in real time, even under network or resource constraints, thereby contributing to safer, more efficient connected environments.
What would you say could be the next big thing in your area of research?
The next big thing in my field will be the development of more efficient and intelligent edge–cloud systems that could support real-time decision-making for autonomous and connected technologies. As computing and communication technologies continue to converge, we will see smarter, faster, and more adaptive systems that can process complex sensor data with minimal delay. These advances will not only enhance the reliability of autonomous vehicles but also enable broader applications in robotics, smart cities, and intelligent transportation networks.
More and more students find themselves training data-driven high-capacity models (e.g., LLMs/VLMs, etc.) in research. What suggestions do you have in terms of the hardware platform - standalone machines vs remote cluster services, choice of graphic cards, cooling systems, and power supplies, etc?
The optimal hardware strategy for training high-capacity models involves balancing fast local iteration with large-scale processing power. Standalone machines are ideal for initial prototyping and debugging, while remote cluster services, such as NJIT’s High-Performance Computing (HPC) platform, are essential for distributed or long-duration training. A practical approach is to use local workstations for quick tests and clusters for extended experiments.
When selecting components, GPUs with ample VRAM—such as NVIDIA’s RTX or A-series cards—offer strong performance for modern workloads. Always pair them with high-efficiency (Platinum/Titanium) power supplies and reliable air or liquid cooling systems to maintain stability during prolonged runs. Understanding how memory, power, and thermal dynamics interact is as crucial as understanding the model itself. Beyond hardware, maintaining organized data practices, collaborative workflows, and ethical awareness ensures that research remains both efficient and responsible.
You hang out in the gym regularly. Please share some insight about how physical health could affect your study and research.
Regular gym sessions play a vital role in maintaining focus and resilience during long research hours. I’ve found that consistent physical activity improves not only physical strength but also mental clarity, creativity, and stress management. Research often involves extended periods of coding, experimentation, and writing; staying active helps sustain energy and discipline. In many ways, maintaining physical fitness parallels the research journey—it requires consistency, patience, and progressive improvement.