Using AI to Help TB Patients in India: My Internship Experience at Wadhwani AI

Mukul Kumar
8 min readAug 13, 2020

--

Image source: https://m.hindustantimes.com/rf/image_size_1200x900/HT/p2/2020/04/14/Pictures/_424c7808-7e59-11ea-aedf-4d2519fcedc3.PNG

This is the moment when I knew, I was not going to Mumbai this year.

Didn’t understand? Let’s travel back to 13th July 2019. (Trust me, I won’t mess anything like people do in DARK)

13July 2019 It was my birthday, but a little different than the usual one. Usually, I spend my birthday like most people do, talking to friends, attending relatives’ calls, saying thank you to the people whom I don’t talk to the rest of the year and spend the evening with my family either going out or arranging a small party at home. So what was unusual? This was the day something big was going to happen to me. I had an online test for the research internship at Wadhwani AI in the evening. I was still not over with the heart-break of India losing the CWC 2019 semi-final followed by the rejection from Goldman Sachs (not to mention the failures from other companies). But here I was revising Bayes theorem, Probability theory, Linear Algebra and Machine Learning on my birthday.

Here we will take a little discourse to introduce Wadhwani AI, it is India’s first independent, non-profit AI research institute developing AI solutions for social good. Apart from their work, I was attracted to their team, which consists of scientists, entrepreneurs, innovators from institutions like Stanford, Microsoft, Google, BCG, IITs and more.

Image source: https://www.wadhwaniai.org/wp-content/uploads/2019/08/fb-preview.png

After reading about Wadhwani AI, their work and the team, I was really motivated to work with them, so I did not want to waste a single minute that could have kept me from passing this test. The test consisted of MCQs based on Machine Learning, Probability, Statistics and Linear Algebra coupled with coding questions based on data structures and algorithms. The questions were of intermediate difficulty. Any person who understands ML fundamentals, Linear algebra and probability could have easily cracked that test. In my case, I had previously done a few online Machine Learning courses which really helped me in solving the questions. My birthday proved to be lucky for me and I was selected for the interview round scheduled 5 days later.

The Day of Interview

An online video interview was scheduled in the evening with the Chief Innovation Officer of Wadhwani AI, Dr Rahul Panicker and Research Fellow, Namrata Deka also a IIIT Delhi Alumnus. When the interview started I was a bit nervous at first, but the interviewers were really friendly and made me comfortable. It lasted for 30 minutes and they asked me to explain one of my projects, discussed my interests and why I want to join Wadhwani AI along with a few ML and probability questions. After it was over, I had this feeling that it went well however you can never be sure until you get the offer letter :p

5 days later, I was at home planning to take a nap in the noon,but was disturbed by a series of WhatsApp notifications’ sound. That was my friends congratulating me on scoring the Wadhwani AI internship. I checked the official mail just to be sure and told my parents about it. They were really happy to hear this news and here I was already dreaming of going to Mumbai next year, unaware of the things 2020 have planned for us :)

March 2020

Coronavirus cases in India were rising and our country went into a complete lockdown for three weeks. I already had made a list of places to visit in Mumbai and was looking for a place to rent for summers. But there was a spur of corona cases in Mumbai. With these many corona cases in Mumbai, my hopes of chilling at Marine Drive went downhill and finally died. I was even sceptical of whether the internship will happen or not.

Image Source: Created using matplotlib

April 2020

I received an email stating the internship will take place remotely (Woohoo!). All my plans and list of places to visit in Mumbai were now a waste, but I was still happy because the internship was happening.

The Internship Experience

The internship began on 4th May and started with a virtual onboarding session with the HR. Then we had an interaction session with the CEO Dr P. Anandan and CIO Dr Rahul Panicker followed by an introduction with Senior Researchers and Research Fellows. The team was really fun and they made us all comfortable on our first day. After that, we were given a brief about how things at Wadhwani AI work, the work they are doing and projects available to work on for the internship. They broadly work on three areas, namely Cotton Farming, Maternal and Child Health, and Tuberculosis. Although all projects were very interesting and really making an impact on users, I personally found the project on Tuberculosis to be resonating with my interests and skill-sets. All the interns then gave their preferences for the projects and I was assigned the project on Tuberculosis to work on along with a fellow intern Satvik from BITS Pilani. Mihir Kulkarni, Research Fellow who was on a break from his PhD at Princeton, was our mentor for the internship.

The rest of the first week was spent completing formalities, setting up infrastructure for a smooth virtual internship, getting to know more about the problem domain to work on for the internship. When I started the internship, I didn’t have a wifi connection at home and was using a mobile hotspot for work for the first few days, which I realised would not be very comfortable moving forward in the internship. So, I thought about installing broadband, and with the ongoing pandemic, it was tough to find one. Fortunately, I found a local broadband vendor and after requesting him several times, I finally got WiFi after a week. (Pro-tip: A smooth WiFi connection will help a lot and is a must for work from home).

My Internship Project — Non-Adherence Among TB Patients

As I told you earlier I was working on a project related to Tuberculosis, which is a disease caused by bacteria and mostly affects a person’s lungs. In 2019, India saw a total of 2.4 million TB cases and around 80k deaths (which is greater than current Coronavirus cases and deaths in India, sadly many of us don’t know how deadly this disease is). TB is however curable if one follows proper medication for the designated period, which is usually of 6 months. But due to social stigma associated with it and the side effects caused by the medical dose, patients often don’t adhere to medication and often do not follow the medical regimen which results in ‘Drug-resistant TB’ which increases the chance of the patient dying.

Our objective in the project was to identify patients who were likely to drop-off the medication regimen so that healthcare authorities can provide suitable interventions and help save their lives. Since Wadhwani AI is an official AI partner of Central TB Division of India, we knew that if built well, our solution has the potential to help millions of people affected with TB and that filled us with a sense of responsibility and enthusiasm.

The Work Culture at Wadhwani AI

Although I wasn’t able to go to their office in Mumbai, Slack was our only thing closest to an office :p. Everyone in the team was really fun and helpful. We used to have daily meetings with our mentor Mihir and Ankit, discussing the daily progress and problems, and a weekly meeting with Dr Alpan Raval, Head of Data Science, Wadhwani AI, who was the principal investigator for this Tuberculosis project.

Apart from these meetings, there used to be weekly presentations with the whole Research team, where all the interns used to discuss their weekly progress, and the issues they are facing, and the senior researchers, people working on other projects used to suggest ideas and way out, which was a really good practice. In the discussions, we were encouraged to bring new ideas to solve problems and they were highly valued. The mentors were always there to help. I also liked the diverse team, the domain experts helped me understand the problem better, how things work on the ground in TB healthcare, what a TB patient’s journey is like from identification to initiation of treatment until its end.

Apart from this, I learned a lot about the issues related to healthcare data (high imbalance, privacy issues, missing outcomes etc.) and how to deal with them while building AI/ML solutions.

My Work From Home Experience

Now, let’s come back to my actual office, my home :p. So, work from home has its own pros and cons, pros being you can stay in your PJs all day, no worry of being stuck at traffic in the morning, getting the comfort of home. However, the problem is you don’t know when your day starts and when the EOD is :-p, you keep getting distracted by the urge of using Twitter, your mom calling you when you are in an online meeting or the loud voice of sabjiwale bhaiya (vegetable vendor) coming from outside when you are discussing something important and the struggle of maintaining a good posture while working to avoid back pain really makes it challenging. But it’s the norm now, so we got to adapt. In the beginning, it was a little hard, but then I learned how to make a balance between work and non-work life and sitting straight (at least for some hours :P). Making a schedule for everything including eating lunch or taking a tea break or a power-nap really helped.

Image source: https://www.facebook.com/BuzzFeedAnimals/photos/i-love-my-work-from-home-co-workers-hannah-hillam-for-buzzfeed-comics/1427919327230058/

Two and a half months came to an end, and I got a call from HR in the evening, that I have an interview tomorrow for PPO. Man! I got a panic attack. Interview for a full-time job that too with a notice of fewer than 24 hours. Now, it was time for a college type ‘all-nighter’. I went through my college notes of Machine Learning and Deep Learning and revised Probability and Linear Algebra. Fortunately, all the interviews went well and I received a PPO from them in the evening, two days before my birthday.

College was yet to open for a month, and now I did not have to prepare for the placements. So, I decided to extend my internship for 3 weeks.

As all good things come to an end, the internship ended in the last week of July leaving me with a lot of knowledge, fun memories, and new friends. I want to thank the whole Wadhwani AI team for this wonderful experience and my mentors Mihir and Ankit for guiding me throughout the journey.

I highly encourage folks who are interested in AI/DL research and are enthusiastic towards using their skill to solve some of the problems in healthcare/agriculture to apply for internship and full-time roles at Wadhwani AI.

Feel free to comment or contact me on Linkedin if you have any questions.

Image Source: https://miro.medium.com/max/2800/1*WGFTn4XD_jcUqNB740btcg.jpeg

As for you Mumbai, see you next year. :-)

--

--