The SPECTRUM club is one of the core technical societies of the College of Engineering and Technology, Bhubaneshwar founded in 2015 by Mr.Bikram Keshari Panda(2015 graduate), scrutinizing on the sole motive of Instrumentation and Electronics Engineering branch having a club of its own.
Apart from being involved in a plethora of competitions and technical events, the club decided to start its very own "Alumni Talk Series" called "Chit-Chat with Alumni".
The goal is to bridge the gap between the current and previous batches of our college and increase the awareness of students in terms of work that one does in college, exposure, jobs, higher studies, and of course networking. These interactive sessions are a small shot at helping college students to step out of their fishbowl life.
The first episode was first hosted on the 17th of October, 2020.
The 9th Episode was hosted on 14th February 2021 and the speaker for the afternoon was Dr.Dharanidhar Dang.
Dr.Dharanidhar Dang is a postdoctoral fellow at the Boolean Lab at the University of California, San Diego (UCSD) where he works with Prof.Debashis Sahoo. His research is focused on designing next-generation computing systems for exascale computing tasks such as Cancer prognosis. He has published 20+ papers in notable venues such as IEEE TPDS, ACM DAC, ACM GLSVLSI, IEEE ICCD, and etc. Dr.Dang is an AAI fellow, past Teaching Fellow, and a member of IEEE and ACM. Prior to joining UCSD, he completed his Ph.D. in Computer Engineering under Prof. Rabi Mahapatra from Texas A&M University in 2018. He has won several prestigious national/international competitions such as Intel Embedded Challenge, Schneider Innovation Challenge, and Intel Youth Enterprise. Dr.Dangโs research interests include silicon photonics, neuromorphic computing, computer architecture, network-on-chip, and deep learning, and bioinformatics. Apart from academics, Dr.Dang spends his time dancing, cricket, cooking, and traveling.
In this session, Dr.Dharanidhar discussed Neuromorphic Computing which implements aspects of biological neural networks as analog or digital copies on electronic circuits, and gave a brief insight of Deep Learning with its diversification.
He also highlighted the fact that by 2022 market value of Deep Learning can cross 100 billion dollars and more.
He briefed his research on Deep Learning by explaining the actual scenario of computer performance. He explained this as:
A computer has memory and processor and we want to analyze its performance. But the Deep Learning process is very fast while memory is tiny/low. We are designing an extra fast processor but we don't have a fast memory. Also, there are tons of layers and the CPU interacts with memory in every cycle which makes the task cumbersome. The entire discussion was on reducing the gap of Memory-Processor.
There has been a lot of improvement in the transistor size nowadays. We can find multiple transistors in a single chip. But the problem of adding so many transistors in a single chip is leakage power(which is more). Active power is the power used to do computing. But here we are trying to show the world an extremely powerful chip with a lot of transistors that consume more power. This contradicts our goal, thus we have to think beyond C-mos.
Next Dr.Dharanidhar explained the process of neuromorphic computing and talked about VLSI systems containing analog circuits mimicking human brain behavior. He compared the difference between how a human brain processes data compared to circuit processing. A human brain is able to process way more amount of information consuming very less amount of power, which is way lesser than how a CPU processes data.
So he came up with the idea of Memristor , which is fast and energy-efficient and does in-memory computing. Basically, he tried to merge the power of data storage and processing into a single chip, i.e., memory and process together, to do very fast in-memory computing, consuming comparatively very lesser power than our today's state-of-the-art mechanisms.
He adapted the technology of Silicon Photonics for the same. Silicon photonics is an evolving technology in which data is transferred among computer chips by optical rays. This enables the data to travel much faster and in more volume than electrical conductors, which is how data is transferred in today's world (CPU routing).
Silicon photonic systems consume 6 times lesser energy and are 12 times faster compared to today's electrical interconnection data transfer units. Then he explained how silicon photonic architecture works and how he implements the concept of deep learning to make the chips more power-efficient to make them consume as much power as required, and not more.
Towards the end of the session, he shared glimpses of his journey so far. Dr.Dharanidhar belonged to a village named "Dangpali" , which is situated on the banks of river Mahanadi, in Bargarh district of Odisha. The village "Dangpali" was named after his great grandfather who was a renowned person in their village. He has done his B.Tech from College of Engineering and Technology, Bhubaneshwar from 2006 to 2010. After passing out from college he was involved in his research at Intel, Bangalore(2011-2012). He then completed his Ph.D. in Computer Engineering from Texas A&M University (2012-2018). He is currently doing his post-doctorate at the Boolean Lab at the University of California, San Diego (UCSD).
He described his fad for Cricket and is known to be a great bowler. He shocked us with his dancing moves. He also added that he has been dancing from his childhood and assured us once he comes down to India he will shock us with his research, his journey, and his moves.
Many attendees got clarified about publishing a research paper, studying abroad, and cracking GRE exam.
We learned a lot from his session and wish him all the best for future endeavors.
Top comments (0)