None of us had anticipated a summer like this. While the world continues to fight an endless battle with this pandemic, I try to look at silver linings within these times of uncertainties. Not only did I have an opportunity to spend time at home with my family, but I also managed to gain some brownie points for building my expertise on household chores.
Hi! I am Sumanyu Ghoshal, a (to be) thirdie at the Department of Computer Science and Engineering. I was a Summer Data Scientist at Prodigal Tech. The past few months were enriching as the lessons learned in those times are ones that I’ll carry with me for years to come.
Here are a few thoughts from my experience:
Why choose data science (and Deep Learning in particular)?
Very early in college, I was a part of a hands-on technical project in the domain of Computer Vision. That’s when I started exploring machine learning and deep learning in particular. This is how my interest in data science grew, as I kept on taking up online courses and reading up tutorial blogs. The course CS231n: Convolutional Neural Networks for Visual Recognition, from Stanford, introduced me to Deep Learning. From there, I’ve kept looking up research papers and other resources to see how problems in this domain are solved.
A solid internship to obtain strong exposure to software engineering and/or data science in the industry was my top priority. Being just a sophomore, it would have been tough to get this exposure in a large MNC. With these thoughts in mind, I gravitated towards startups. Thus, I began searching for internships through the PT Cell as well as LinkedIn contacts and referrals. In this process, I found Shantanu Gangal’s profile on Linkedin. That’s how I got introduced to Prodigal.
For those who are interested – Prodigal is a Y-Combinator backed startup, headquartered in Silicon Valley, which provides banks and lenders AI-based solutions to improve their revenue yield and reduce compliance risk. Prodigal essentially leverages AI to bring efficiency in the debt collections industry.
The opportunity to work with an exceptional team of engineers and data scientists led by IITB CSE alumni co-founders, Shantanu Gangal and Sangram Raje, was incredibly alluring to me. I felt no hesitation in getting in touch with them to seek an internship, as I strongly felt that joining them would end up giving me a great learning experience.
Within a few weeks, I had an interview with Sangram, wherein I was apprised of the available opportunities and where they could fit me in. By the end of the day, I received a confirmation of my selection. An internship at the end of my sophomore year, which could substantiate a product to be used in the real world, was incredibly appealing. I happily accepted the offer.
By mid-March, the world had almost come to a halt with the onset of Covid-19. When the institute decided to go ahead with preponing the vacations, I wrote an email to Neil Shah, the chief of staff, explaining my situation, and within a day, they happily agreed to bring me in to start working on the project!
I worked on the speech analytics RnD team of Prodigal. I would also occasionally participate in conversations with the Data analysis team.
The project that I primarily worked on is Voice Biometrics. As the name suggests, the aim of the project is to create a solution to verify on any sensitive phone call, whether or not the correct person is speaking within 3-5 seconds, using just the speech of the person. From the very beginning, I was allowed to take up the responsibilities to get the various subtasks done within reasonable deadlines.
The project involved the following phases: (For the Tech Enthu :P)
- Research: The project began with reading up some research papers and filtering out some state of the art and ‘realistic’ deep learning models that could be used to solve the problem at hand. With the available speech data, we gravitated towards a Similarity generator for two audio clips using Siamese Convolutional Neural Networks. As far as extracting features from the audio clips is concerned, I stuck to the use of Mel-Frequency Cepstrum Coefficients (MFCCs) based on some research papers and my own experimentation.
- Preparation of the Dataset: This phase required going through the available data and figuring out ways to clean them. This involved working on different types of “noises” which people find in phone recordings. For example, cross-talk on phone recordings ie. a channel recording having the speech and noise from a different channel, is a problem in many open-source phone call clips. Unaware of such issues before going through the dataset, this phase made me realise about the constraints that a real-world data science project would face and the kind work that goes into the pre-training phase.
- Training the Model: This process is used to make the model ‘learn’ how to make the correct decisions. I went ahead with training various examples of this model and tried out some novel adjustments and techniques that are used to improve the accuracy of a CNN model. Training deep learning networks takes time, and with some different set of models that I had made, this phase did take up several days.
- Deployment: This was a slight detour from the Data-Science aspect of the project. The deployment of the model itself was a very different experience, as I got to explore Amazon Web Services products and try out serverless product development, something which I think an undergrad wouldn’t be exposed to unless having worked in the industry. Serverless applications are essentially based on the backend-as-a-service provided by cloud computing platforms like AWS, Microsoft Azure, and Google cloud. The constraints in space and RAM for the deployments were some of the issues that I needed to work on for the deployment.
- Testing: After the deployment of the model, we went ahead with testing the interim product on their database. This gave the RnD team a direction to the work that needs to be done to get the project in the production phase.
- Client-Wise Fine-Tuning: Using our clients’ data, I was involved in fine-tuning the model for each client using transfer learning methods, in order to make the model as accurate as possible, and fine-tuned for each client.
I would have presentations on a fortnightly basis, which anyone from the company could join to get to know more about the project, ask questions related to the progress and brainstorm different cases and issues. These presentations were the incentive for me to keep working on this project with conviction, as these would act as a great way for getting a direction on how to go ahead with the project.
Working From Home
Since I haven’t had the opportunity of working in an office environment, it would be difficult for me to comment on the experience as compared to working at an office.
The constant interactions with Sangram and the rest of the RnD Team to resolve my concerns, review the progress of the project and provide feedback and suggestions on how to go ahead, made me feel connected to the team. On the company-wide front, with the monthly all-hands meets and fun surveys, there was a whole-hearted effort to make sure that everyone’s morale was high while the pandemic lived with us.
The primary challenge that I faced while working at home was to stay focused on the same project for a longer time than I am accustomed to. Another issue that we all face and must become accustomed to during this pandemic is maintaining a balance between work-life and daily chores.
The one thing that I missed out on is to observe an office environment in a startup, an experience that would’ve been really exciting.
The experience that I had this summer is something that I truly consider fulfilling and enriching. I would’ve definitely enjoyed working at an office, but given the situation we all are in, I find myself blessed with the opportunity. From working on the technical aspects of a data science project to presenting my work for both the Technical-minded and non-technical minded groups, the experience was definitely an enriching one. I would definitely recommend people to take up opportunities at startups, especially if the company and the individual’s interests are in tandem.