Saikiran Bulusu – Where signal processing meets academic rigour and military precision

Dr. Saikiran Bulusu, an assistant professor at IIIT Hyderabad’s Signal Processing and Communications Research Center (SPCRC), discusses interesting research in machine learning and signal processing. Army life and love of mathematics taught him discipline, his Ph. D showed him humility, and teaching continues to add new dimensions to his study.

An M.Tech. from IIT Madras, Dr. Saikiran Bulusu, worked at Samsung Research briefly before moving to IISc Bangalore as a research associate. He completed his Ph.D. in Electrical Engineering and Computer Science from Syracuse University and a postdoctoral study at The Ohio State University. His research areas are Federated Learning, Large Language Models, Robust Statistics, Compressed Sensing, and Group Testing.

“The emphasis at IIIT Hyderabad is to build on one’s education and training to frame a problem, solve it, and take it from theory to a working prototype. This pipeline from education to research to enterprise is what inspired me to join the Institute in May 2025”, observes Bulusu, who teaches courses in machine learning and wireless communications at IIIT-H.

Military discipline meets academic rigour
“Being an army kid meant that every posting was an upheaval”, he recalls. “But it taught me adaptability and the ability to connect with people quickly”. Early education was spread across Kendriya Vidyalayas and Army schools, from far-flung Faridkot, Ladakh, instilled resilience early on. He later pursued B.Tech. in Electronics and Communication Engineering from Mahatma Gandhi Institute of Technology, Hyderabad, where his interest in signal processing began to take shape. “That phase gave me a simple but enduring belief—knowledge is power.”

Earning his stripes at IIT and IISc
While Electronics was his initial choice, he soon realized that his interest was in signal processing and communication.  His M.Tech. at IIT Madras broadened his worldview, as he interacted with a wide spectrum of researchers from around the country.  “IIT was a different ballgame that completely humbled me”, he admits. “I had pegged myself as a good student till I met the geniuses at play there. I caught the research bug while working on my thesis on Schemes for quickest change detection, based on on-off observation control and CUSUM with adaptive censoring of sensor networks, under Dr. SriKrishna Bhashyam.”

With the intent of sampling corporate life, Bulusu joined Samsung R&D Institute as Lead Engineer, working with Bangalore and South Korea on system design of baseband functionalities and DSP firmware development for LTE Modem Chipsets.

Two and a half years later, an appetite for research brought him to IISc Bangalore as a project associate, where he worked on device-to-device communications in Cellular Systems. “Those three years were a big learning curve for me”, he reveals. “I learnt the value of knowledge, mathematical rigour, and chiefly how to approach research, from my advisor Dr. Neelesh Mehta.” A paper on wireless communication that was rejected thrice before it was published were teachable moments in the art of persistence and staying motivated in the face of rejection. ‘One must continue striving without hope and without fear’ is a quote famously ascribed to Bhagat Singh, which became his mantra.  

Syracuse, PhD and a world vista
Bulusu moved to Syracuse University for his PhD in Electrical and Computer Engineering (2017-2023). From 2G, 3G and 4G wireless communication at Samsung and IISC, he switched to machine learning at Syracuse.

It was his advisor, Dr. Pramod Varshney, who nudged Bulusu into machine learning, an emerging domain at that time. A purist in terms of signal processing and estimation theory, his advisor had a visceral disdain for machine learning models, which he termed ‘black boxes’. For him, the reigning question was; can we deconstruct the black box and help people understand what it is doing? Can we use tools of signal processing and estimation theory to explain this black box? He elaborates, “I spent two years solving this problem, which translated into my first paper during my PhD I also worked on signal processing problems, specifically on compressed sensing, often used in image processing, wireless communications, and astronomy”.

His six-year doctoral tenure coincided with the pandemic lockdowns. Being sequestered in their accommodation at Syracuse turned out to be an interesting time for Bulusu and his roommates. By the time the world opened up again, they had learnt to whip up Momos and Chinese cuisine, and bake cakes and breads, apart from soul food like dosa and sambar.

On the academic front, new opportunities turned up for online research collaborations across the world. “I had great advisors in Dr. Pramod Varshney for strategy and Dr. Venkata Gandikota for tactics, who guided my thesis on Robust Machine Learning in the Presence of Adversaries”. Interacting with multiple nationalities and learning about different socio-cultural perspectives, from Romania, France, the USA, Rwanda, China, South Korea, and Mexico, broadened his research horizon.

Fun Post-doc at Ohio and a federated solution
When he moved to The Ohio State University for his postdoctoral work, Bulusu worked with great academics who shaped him as a researcher and opened up new avenues for him to explore. “Beyond theory, I was working on real-world problems that had a practical impact. Most universities were facing a resource crunch while training large models. The major problem is that all computing resources are monopolized by big corporations (Google, Meta, Microsoft et al) who have high-end GPUs to train large models, which academia cannot afford. We looked at democratizing this model.  Can we use lower-end GPUs to train larger models by breaking these models into smaller chunks and training them collaboratively, across these lower-end GPUs? We demonstrated that it can be done, and that was a big part of my postdoctoral work”, says Bulusu, who published papers on orthogonal fields, Machine learning and group testing. He explains that Group testing identifies a small number of “defective” items within a large set by testing groups rather than individuals. It focuses on designing efficient testing strategies and decoding methods to minimize the number of tests while maintaining accuracy.

“I also worked on robustness and explainability. My interest was in the reliability of these machine learning models, particularly when an adversary corrupts the model itself. How can we ensure that the model is robust against such chaos and corruption?  In the explainability part, can you explain how a model is generating the output it is giving?” Machine learning conferences rank high in his estimation, for the interesting networking platforms they offer for researchers. “I got to interact with researchers whose work I’ve been following, and that was fun. Moreover, ISIT, while it is a smaller group, I particularly loved the fact that I could have one-on-one conversations with researchers. I have also been part of the organizing committees at ICAASP, NCC, and internal workshops here at IIIT-H”.

Experiential immersion into the thriving IIIT-H ecosystem
In February 2025, when Bulusu was considering his return to India, he was offered a faculty role in IIIT Hyderabad.

“The whole IIIT-H experience has been wonderful. I chose IIIT-H for the flexibility it offers. They do not subscribe to the conventional notion of departments. At IIIT Hyderabad’s Centers, there is plenty of interdisciplinary work happening between faculties and centers “, notes Bulusu who is working with the Smart City Research Center, building machine learning algorithms for crowd monitoring and management. “I have collaborations with SPCRC faculty on IoT-related problems and wireless communications, and MLL faculty on federated learning and trustworthiness”.

That which built grit and valour
“I grew up listening to Dad narrating stories from his training days at NDA and IMA; running a full battle proficiency test immediately after recovering from double fractures on his legs. His stories positively influenced me to be adamant and determined”, muses Bulusu.  

An experimental music buff, the scholar’s listening choices swing wildly from classical, jazz, blues, math rock and psychedelic rock, to ghazals, Bollywood, Tollywood, and Persian tunes. Reading choices are classically sober in George Orwell, Dostoevsky, and Tolstoy.  

Outdoor sports and fitness choices tend towards squash, running, bouldering, and skiing when the landscape permits. He enjoyed hiking and road trips around Karnataka, Canada and the East Coast of USA.  Bulusu loves hanging out with his circle of friends; from the city and from his undergrad, masters and Ph.D. days. “I do spend a lot of time with my family, whenever I can”, he adds.

In the future, Dr. Saikiran Bulusu hopes to expand his work in Trustworthy AI. “I am particularly interested in how properties like robustness, fairness, privacy, and explainability interact,” he says. “What happens when we try to optimize for several of them simultaneously?”  It may be a long road ahead, but the journey promises to be exciting and edgy, given the caliber of his research ecosystem.