Sailaja Rajanala is a postdoctoral researcher at Monash University Malaysia. 

She completed her PhD at the Indian Institute of Technology Hyderabad. Her thesis was building context-based recommender systems in e-commerce, scientific research articles, and social networks.

INDIAai interviewed Sailaja Rajanala to get her perspective on AI.

How did you get started with artificial intelligence? Could you tell us about your academic background?

During my bachelor's, I liked attending and presenting technical sessions. In one such session, I gave some recent developments in HCI (Human-computer interaction). At the end of my talk, one of the attending lecturers asked me how a machine or robot would accomplish a new task. I confidently said that it would not unless and until we fed the instructions into it. The lecturer, however, argued that it was not the case and that machines can learn to perform tasks. Although I was intrigued by the idea of a machine's capability to learn things by itself, being naive, I was still not convinced of how this was possible. So I decided to study the study of machines, i.e. AI as we know it now. 

About my academic background, I completed my bachelor's in computer science in 2013. My final year project was a multilingual text translator named "SwaGirAnthar", meaning translating(anthar) to our(Swa) language(Gir). The idea was to propose a translation tool as a backend to a speech processing software that could translate multilingual speech to either Hindi or Telugu because humans tend to mix up multiple languages when they speak. 

After my undergrad, I took the position of Assistant Software Engineer in Informatica at Accenture Services Pvt. Ltd from 2013-2014. As a part, I worked on a banking firm's data mining and warehousing processes. 

In July 2015, I was selected as a direct PhD candidate at the Indian Institute of Technology Hyderabad. 

What were your initial challenges in the field? How did you overcome them?

My initial challenge would have to be of understanding the process of conducting research. Up until my bachelor's, proposing a solution was all about coding. It was during my PhD that I understood the real challenges of research. How do we have to have good motivation for a problem? Most of all, I must maintain a balance between reading literature and working on it. There is a vast sea of research out there, and it is easy to get carried away and demotivated thinking about what new you can contribute. Similar to the problem of finding a balance between exploration and exploitation in Reinforcement Learning. Of course, I had my advisor, Dr Manish Singh, thank him for it. He helped me understand when to stop foraging and start acting on the problem. 

According to a detailed Market Research Future (MRFR) "Natural Language Processing Market research report," the market will be worth about USD 341.7 Billion by the end of 2030. How should universities and students prepare for this growth to handle it?

Ironically, they posed this question to me during my bachelor's, and I never understood it. I was told that data is everything and that there will be an explosion of data to deal with in a few years. When I look back, the students and universities have adapted to this boom by offering focused streams of courses in data mining, data science, etc. 

In addition, I feel there must be a more outstanding collaborative research network within and outside the country. The collaboration will help us reduce the redundant part of the research cycle while contributing more to the actual problems. Moreover, collaborations promote the free flow of knowledge. 

You completed your PhD in India and are now pursuing a postdoctoral position in Malaysia. What differences do you observe between the two nations?

The working culture in Malaysia and Monash is slightly different because it is more diverse. Monash is spread across Australia, Europe and Asia because we work with people from other countries and cultures. It is an enjoyable learning experience, both professionally and personally, to interact with and work with such a diverse community. 

How did you deal with journal article rejections in the beginning?

I talked to my peers, seniors, and advisor and realized that I was not alone in this process. It helped me to boost my confidence and to look at the positives of the feedback given by the reviewers. 

Could you tell us about your doctoral research?

Understanding the context in the recommender framework was the central theme of my doctoral thesis. The context controls how the end user may find content (not)relevant. At the early stage of my thesis, I leveraged the context mainly from the user knowledge base. Later, I emphasized learning from a limited context that includes recommending academic venues to researchers based on only the title and abstract of their work. According to the CORE ranking, my present work was published in SIGIR'22, an A* conference. It goes beyond the correlational nature of context and explores causality. Furthermore, it goes beyond explainability while learning from limited information. 

We all say that machine learning and deep LearningLearning are new technologies that have changed the world. But we often don't notice the unique problems they bring with them. One of them is how these fast technologies affect the environment. What do you think about that?

I understand that deep learning requires enormous computational power and devices that are not very environmentally friendly. So was the case when we saw the advent of computers and the internet. But the internet and computer have helped us digitalize the world, thus significantly reducing the dependency on paper and travel. Moreover, companies are already investing in using deep learning to design energy-efficient spaces by deploying smart gadgets that can detect and arrange power on a need basis. So I feel that the benefits of deep learning will soon outweigh its disadvantages. 

What advice do you have for people who want to work in the field of research into artificial intelligence? What should they concentrate on to advance?

My advice would be not to consider AI as a standalone subject. There are prerequisites to understanding and appreciating AI. Concentrate on the foundations like calculus, statistics, probability, linear algebra, optimization and programming. 

Learn to strike a balance between research and programming. First, start with research, establish the problem and proceed to the program. Value quality over quantity, and try not to fall in the race of publishing by compromising the research quality.  

Look for lesser-explored research domains instead of publishing on a trending research topic. For instance, in my opinion, Trustworthy AI has a promising future in AI. I have witnessed people working on real-time applications emphasize accountability and certainty. In addition, trustworthy AI ensures that AI systems are privacy-preserving, explainable, fair etc. 

Most of all, collaborate. Talk to your friends within and outside the department, your seniors, juniors and professors. Communication will help you gather newer insights and widen your research horizons.

Could you name some important research papers and books that have changed your life?

I appreciate multidisciplinary works that enhance the research of a field in collaboration with another area. So be it the discovery of the double-helix structure of the DNA by a group of chemists, biochemists and physicists. Likewise, the first-ever picture of a black hole was only possible through the combined efforts of computer scientists, mathematicians, and astrophysicists. 

I also suggest reading some of the very early papers in machine learning, as early as the 1980s, like the first CNN described in Kunihiko Fukushima's work "Neocognitron: A Self-organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in Position". These works are so influential that they transport you to the thoughts of the researchers at the time. It is fascinating how they believed machine learning would be a valuable asset to the world.