Narendra N P works at RingCentral as a data scientist. He is primarily an AI researcher with a non-IT background. He earned his master’s and doctoral degrees from the Indian Institute of Technology, Kharagpur.
He completed a postdoctoral research fellowship at Aalto University. His research interests include speech processing, machine learning, deep learning, speech synthesis, signal processing, natural language processing and Paralinguistic speech processing.
INDIAai recently spoke with him to learn about his research journey and better understand his work and vision in speech processing.
From electrical and electronic communication engineering to speech processing and synthesis research. How did this transition occur?
The transition was smooth and adaptable, as I learned the basics (such as digital signal processing) required to work on speech processing research in Electronics and communication engineering.
What motivated you to pursue a career in speech processing and synthesis?
I was interested in speech and image processing right from my Engineering days. While looking for my Master’s thesis, I got a chance to work on speech processing. My supervisor (Prof. K. S. Rao from IIT-Kharagpur) motivated me to pursue my research in speech processing.
It’s great to know that you were involved in the project development of text-to-speech (TTS) synthesis in Indian Languages. Can you tell us about some of the exciting challenges you faced while working on this project?
There are many challenges at different stages of project development. From scratch, we developed a text-to-speech (TTS) synthesis system in the Indian language (Bengali). The data collection was a challenge as there was no previous data for TTS. Developing a good quality speech on par with other universities was a challenge. After creating the TTS system, we wanted to train visually impaired individuals on how to use it. Training visually challenged people was also a challenge, as we did not have any experience with such a task. In the end, challenges are always a good learning opportunity.
The Global Conversational AI Market was worth USD 8.24 billion in 2021 and will reach USD 32.30 billion by 2028, growing at a 21.5 per cent compound annual growth rate between 2022 and 2028. How far has India progressed in the field of conversational AI?
In my opinion, India is progressing at an excellent pace to match up with the global market. The number of start-ups focusing on conversational AI has increased due to this progress. Most companies have an AI team trying to solve the problems from an AI perspective. Nowadays, there are features related to conversational AI in most virtual meetings and customer feedback services. The number of offered conversational AI features is known to increase in the coming years.
What is your job as a data scientist in your organization?
I am mainly involved in the research and development of one of the problems related to the broader area of conversational AI.
What was the most challenging part of your research, from problem identification to data collection?
The problem identification is the first and most challenging part. Once a problem is identified correctly, the rest will fall into place.
For people who don’t have a background in IT, what should they do to get ready for a career in AI?
First, people should understand the basics of AI. Then, understanding the core topics related to AI, such as machine learning, deep learning, and statistics, is crucial. Then, learning programming languages such as Python in AI is also important. Finally, the candidate has to get hands-on experience developing AI models using freely available data online. People can either learn the things mentioned above online tutors or offline from universities.
Who is your role model in AI research, particularly in conversational AI?
My role models are my supervisors (Prof. K. S. Rao from IIT-Kharagpur and Prof. Paavo Alku from Aalto University, Finland), who guided me during my PhD and Postdoc.
Which research articles and books on artificial intelligence inspired you during your research?
I was inspired by “Theory and Problems of Statistics by Murray R. Speigel” and “Machine learning for Hackers by Drew Conway and John Myles White.”
Source: indiaai.gov.in
