Divya Mahajan

Alumna Returns as Faculty Focusing on Sustainable Deep Learning Infrastructure

After receiving her Ph.D. from the School of Computer Science (SCS), Divya Mahajan is back at Georgia Tech, this time as faculty.

Mahajan began this fall as an assistant professor in SCS and the School of Electrical and Computer Engineering (ECE). After finishing her Ph.D. in computer science, Mahajan spent four years at Microsoft designing data centers for artificial intelligence (AI).

Mahajan’s research focuses on building deep learning infrastructure while optimizing for a limited carbon footprint. That her research has a positive societal impact is important for Mahajan. She said this is one of many factors that influenced her return to academia.

“As computing researchers, we can have an impact by building systems and infrastructure with sustainability as a metric. We have this opportunity to be able to directly think about climate change, especially as computing becomes such an important area,” she said.

Mentoring and collaborating with students are also part of what drew Mahajan to Georgia Tech.

“Mentoring students closely, seeing them grow, and teaching them to be educated in certain fields- and them teaching you something as well along the way- is unique to academia,” she said.

Divya Mahajan
Divya Mahajan began as an assistant professor at Georgia Tech this fall. Her research focuses on building deep learning infrastructure while optimizing for a limited carbon footprint. (Photos by Terence Rushin/College of Computing)

What interests you about working at Georgia Tech?

Georgia Tech offers a highly cross-disciplinary environment. ECE and SCS both foster a strong collaborative atmosphere. Another very important strength of ours is that students and faculty here are highly motivated, collaborative, and, most importantly, dedicated to the success of each other.

What will your research consist of?

My research will have the overall thrust of building architectural solutions for the end-to-end data processing pipeline. As a first step, my work will take a radical approach towards distributed domain-specialized systems (architectures, device placement, networking) that will enable the next generation of large AI models. The goal is to build these architectures and systems with sustainability and carbon footprint as an optimization metric. Next, I will look at solutions that break away from the traditional CPU-based stack and integrate closely with data collection and management systems.

How did you get interested in this field of research?

I have been working on building domain-specialized architectures for machine learning for about a decade. My Ph.D. dissertation was the first to build full-compute-stack solutions for machine learning. Following that, I spent a few years in the industry at Microsoft, where I had an opportunity to understand the real-world challenges of building such large-scale systems. Now my academic position offers me the opportunity to tackle these challenges from a new perspective, enabling me to design equitable solutions that can achieve a broader impact.

What are you most looking forward to in your new position?

My research journey began here at Georgia Tech, so returning feels like a homecoming. I am thrilled to be back, although with a different role. I understand this place’s strengths and challenges from a student’s perspective and now from a faculty perspective. So, I hope to foster a strong research culture with graduate and undergraduate students and collaborators in my lab. I’m very excited about working with students, teaching, and research mentoring. I will also be working on building a research course on distributed machine learning.