One of the biggest challenges facing researchers is, teaching machines to replicate tasks that require higher order thinking !
Briti Gangopadhyay, our next pathbreaker, AI Researcher at Sony (Tokyo, Japan), develops AI based automatic systems for delivering ad campaigns for Sony’s products.
Briti talks to Shyam Krishnamurthy from The Interview Portal about her PhD at IIT Kharagpur, where she worked on development and verification of reinforcement learning algorithms with a focus on safety during autonomous driving.
For students, If you strongly believe in your chosen path, things will eventually fall into place.
Briti, can you tell us what were your growing up years like?
I grew up in the quaint town of Ranchi till the age of 12 and then moved to Kolkata. My father is an architect and my mother was a yoga teacher. During my childhood my mother was adamant in making me capable of self defense. Thanks to her I started learning Karate at the age of four and completed my black belt by the age of 12. My mother also exposed me to other cultural learnings like Bharatnatyam, poetry and singing. The experience from varied extra curricular activities helped shape my adventurous spirit and curious mind.
What did you do for graduation/post graduation?
I studied computer science and engineering for my graduation. For my post graduation I studied embedded controls and software from IIT Kharagpur.
I also did my PhD in Reinforcement Learning from IIT KGP.
What were some of the key influences that led you to such an offbeat, unconventional and cool career Applied AI ?
The key factor in choosing a research career was to experience the joy of creating something new.
I was fortunate to have exceptional mentors who shaped my research interests. My first internship was under Sandeep Patil, now CTO of IBM ISDL, India who helped me gain a new perspective towards looking at an industry problem. Then, I had the opportunity to intern under Prof Siddharth Khastgir for University of Warwick who laid the foundation stones for my research career. Finally Prof. Pallab Dasgupta my PhD supervisor who has been instrumental in shaping me as a researcher.
One of the events that convinced me to take up research was a research internship at University of Warwick during summer 2018. The internship gave me the flavor of working with interesting future technologies like self-driving cars. This fascinated me to pursue this field further.
A turning point in my career was appearing for and clearing gate examinations which then opened further avenues for a research career.
How did you plan the steps to get into the career you wanted? Or how did you make a transition to a new career? Tell us about your career path
I started working as a Salesforce developer after my graduation for Persistent Systems. We were building a health care system for a US client. While the excitement of entering the task force was high in the beginning it started dying down as the tasks started becoming mundane. I appeared for the air force selection board while doing the job where the interviewer told me I should go for IIT. Taking his advice into consideration I started preparing for GATE.
On cracking GATE, I travelled through the length and breadth of India giving interviews at different IITs. Finally I received a call from IIT Kharagpur from Embedded controls and software branch. There has been no looking back since then. I was fortunate to study under and with some of the brightest minds of the country. They taught me the subject so well that I topped the course and was awarded a silver medal. During my master’s, I did 3 internships.
One was with an in-house IIT Kharagpur startup Mbreath. The startup was building a sleep monitoring system and I was given the opportunity of building an efficient storage system for their application.
During summer I did an “Extreme Blue” internship at IBM ISDL (India Systems Development Lab) where we worked with deep learning for data geofencing.
Extreme Blue internship is a coveted internship program launched by IBM where students from top universities are given an opportunity to pursue a 2 months internship at IBM labs under the mentorship of IBM researchers. The project that we worked on involved classification of personal data (like credit card) and non personal data in their storage system using deep learning. The access to the personal data was then required to be restricted based on geolocation. For example, some private documents could only be accessed from Europe under global law. This selective access is termed as geofencing.
This was a great learning opportunity on real projects. I then moved to the UK for 3 months interning with Warwick manufacturing group creating test cases for autonomous driving systems. This was exciting as we were working on emerging technologies.
After my masters, I stayed in IIT KGP to pursue PhD under the supervision of Prof. Pallab Dasgupta. I spent 4 wonderful years in the Formal methods lab inventing techniques for making reinforcement learning safe. Being in one of the biggest campuses, I also had the opportunity to pursue my other interests like sports. I would regularly go swimming, play squash and participate in inter college general championships. This helped me grow holistically.
My PhD research interests were definitely influenced by my time at Warwick. I was fascinated with self driving cars.
My scholarship was funded by TCS for my PhD. As a part of this program, I was required to present my work to TCS scientists. This helped me gain good feedback. However, I did not need to work on problems provided by them and could do independent research on topic of my choice.
Reinforcement learning (RL) is a machine learning technique that aims to learn human like behaviour by gaining feedback from the environment. For example, when a baby learns to walk, he may fall down and get hurt and learn that it is a bad action. If he stands up, his mother will clap teaching him it is a good behaviour. RL exactly tries to mimic this through reward signalling. For teaching the machine how to drive automatically we use RL with rewards that that can capture human like understanding like crashing is bad and driving at desired speeds is good. My work specifically revolved around how to make these intelligent driving policies safe for real world use.
On completing my PhD, I wanted to apply my knowledge to real world business applications. With this vision I joined Sony R&D in Tokyo where I have been working since the past one year.
How did you get your first break?
My first break in the research field came with going for a research internship at University of Warwick UK. Here I got a taste of working with futuristic technologies like autonomous driving. This opportunity was provided by my Master’s supervisor Prof. Dasgupta under whom I also pursued my PhD.
After my PhD, I got to know about a job opening at Sony when I was attending a reputed international conference for presenting my work. I applied directly through Sony’s job portal. During interview, I had to explain my research work among other technical questions. A good research profile during PhD helped me secure this opportunity.
What were some of the challenges you faced? How did you address them?
Challenge 1: Preparing for GATE while having a full time job was challenging. I overcame this with perseverance and dedication. I would come back from work at 8 PM, then study till 4 AM in the morning and then sleep 2 hrs. I would go swimming at 6 AM and then to the office. I don’t always recommend this lifestyle but it is sometimes required if you want to achieve something in life.
Challenge 2: I was the first student to start reinforcement learning research in my lab. There were a lot of problems to solve, like resources and gaining domain knowledge. My passion for research helped overcome these challenges. If you believe in your chosen path, things will eventually fall into place.
Challenge 3: Research in general is hard because you are trying to create something that does not exist and convince others that your method is correct. 90% of the time things won’t work. Resilience towards failures helps overcome these challenges.
Where do you work now? What problems do you solve?
I work for Sony R&D in Tokyo as an Artificial Intelligence researcher. I develop automatic systems for delivering ad campaigns for Sony’s products. This requires analytical and problem solving skills and knowledge about latest technology trends. Some of the research skills required were acquired from PhD. The others are acquired through continuous learning and reading technology blogs. A typical day in this job involves understanding the business problems, developing techniques to solve them and building prototypes.
What do you love about your work?
The job lets me interact with researchers from all over the world. I work with a dynamic team where I can learn everyday from fantastic engineers and researchers. It can be a little intimidating at times but growth happens when you are a bit uncomfortable.
How does your work benefit society?
My work benefits society by pushing the boundaries of science and helping humans automate mundane tasks.
Tell us an example of a specific memorable work you did that is very close to you!
During the start of my PhD, I designed a system that could automatically drive cars using AI while maintaining road safety. It was a fun project as I could play around with the cars in a simulator while learning driving policies for them. Later the work got published in a top journal.
Your advice to students based on your experience?
Chase your dreams and never give up. Also, there is no one dream that you will be chasing in your lifetime. Dreams change and that is okay. You achieve something and then you would like to do something else. Go towards a career that you enjoy, only then you will be able to excel at it.
Future Plans?
I would like to give back to my country by pursuing research in India.