Memory Device Technology for AI/ML Computing Course | IIT Hyderabad

Course Details

Memory Device Technology for AI/ML Computing: Bridging Hardware and Intelligence

The rapid ascent of Artificial Intelligence and Machine Learning (AI/ML) is not just a software revolution; it's a hardware one. At its core lies a critical challenge: the "memory wall." Traditional computing architectures, built for logic and deterministic tasks, struggle under the massive, data-intensive workloads of modern AI. This is where memory device technology becomes the pivotal battleground for the future of efficient computing.

Recognizing this need, the Indian Institute of Technology Hyderabad offers a seminal 12-week course, "Memory Device Technology for AI/ML Computing," led by the distinguished Prof. Shubhadeep Bhattacharjee. This course provides a deep, structured exploration of the memory technologies that power today's systems and will define tomorrow's intelligent machines.

Meet the Instructor: Prof. Shubhadeep Bhattacharjee

Prof. Bhattacharjee brings a rare blend of industrial and cutting-edge academic research experience to the classroom. His journey from BITS Pilani to ST Microelectronics, and then to pioneering research at IISc Bangalore (where he received the Institute Gold Medal for his PhD), Intel Labs, and the University of Manchester, has centered on next-generation electronic devices. His work on sub-thermionic 2D transistors, neuromorphic devices, and quantum devices in 2D moiré superlattices positions him uniquely to guide students through the frontier of memory technology for low-power AI computing.

Who Should Take This Course?

This course is meticulously designed for a targeted audience:

• Undergraduate Students (3rd/4th Year) in EE and ECE streams with an interest in Semiconductor Devices and VLSI Design.
• MTech and PhD Students actively working in Memory Device Technology, Neuromorphic Devices, and Hardware for AI/ML.

Prerequisites: A foundational UG course in semiconductor devices and electrical circuits. Knowledge of analog circuits is beneficial but not mandatory.

Industry Relevance: The course content is directly aligned with industry needs. It has received support and guest lectures from Micron Inc, and is highly relevant for roles at leading semiconductor firms like GlobalFoundries, Intel, and TSMC. Past students have noted its significant value for campus placements.

Course Overview: A Three-Part Journey

The 12-week curriculum is structured to build knowledge from the ground up:

Part 1: Foundations of Modern Memory Systems (Weeks 1-6)
We begin by understanding the ecosystem. The course delves into the evolving computing paradigms, the Von Neumann bottleneck, and the hierarchical memory-storage organization that makes big data accessible. Students will gain a firm grasp of memory array architecture, peripherals, and the core device physics and operations of workhorse technologies: SRAM, DRAM, and NAND Flash.

Part 2: Emerging Non-Volatile Memories (eNVMs) (Weeks 7-8)
Here, we explore the future of storage-class memory. The course covers the fundamentals and commercialization status of key eNVMs that blur the line between memory and storage: Memristors (RRAM), Phase-Change Memory (PCRAM), and Magnetic RAM (MRAM). These devices are crucial for dense, fast, and low-power storage.

Part 3: Hardware for AI/ML & Beyond (Weeks 9-12)
This is where the course connects memory technology to its ultimate application. We analyze why current hardware is inefficient for AI/ML, introducing the computational and energy costs. The curriculum then pivots to revolutionary architectures:
- Neuromorphic Computing: Mimicking the brain with Leaky Integrate-and-Fire (LIF) neurons, synaptic plasticity (STDP), and Spiking Neural Networks (SNNs).
- In-Memory Computing (IMC): The paradigm-shifting approach of performing computation within the memory array itself, drastically reducing data movement and energy consumption.

Detailed 12-Week Course Layout

Key Textbooks and References

• Yu, S. Semiconductor Memory Devices and Circuits. CRC Press, 2022.
• Wan, Q. & Shi, Y. (Eds.). Neuromorphic Devices for Brain-inspired Computing. Wiley, 2022.
• Additional reference and review articles are provided within the lecture slides.

Conclusion: Why This Course Matters

The transition to an AI-driven world demands a fundamental rethinking of hardware. "Memory Device Technology for AI/ML Computing" is more than just a course; it's a roadmap to the future of computing. By offering a comprehensive journey from established SRAM/DRAM to revolutionary neuromorphic and in-memory computing architectures, it equips the next generation of engineers and researchers with the knowledge to break the memory wall and build the efficient, intelligent systems of tomorrow.

For any student or professional aiming to be at the forefront of semiconductor and AI hardware innovation, this course from IIT Hyderabad is an invaluable investment.

Explore More