Biography
I am working as a Lead Engineer at Samsung R&D Institute India - Bangalore (SRI-B) specializing in the field of Machine Learning and Artificial Intelligence. I joined SRI-B shortly after my Bachelor’s (B.Tech) in Computer Science and Engineering from IIT Kanpur.
Throughout my tenure at SRI-B, I have gained extensive experience across diverse projects, encompassing domains such as Machine Learning, Federated Learning, and Android app development. The majority of my projects lie in the domain of Federated Learning, where I have specialized in training machine learning models within the limited computational capacity of users’ edge devices. By adopting this approach, we have achieved significant advancements in preserving user privacy as their data remains securely on their devices, thus never leaving their possession. Notably, my research contributions within these projects have yielded patents and publications, which are detailed in the “Research Work” section.
Before my involvement with SRI-B, I gained valuable experience through an internship at Hike Pvt. Ltd., located in New Delhi, during the summer of 2017. Within this role, I focused on developing an Image Classifier capable of categorizing images into predefined classes. Apart from that, I actively contributed to the Humanoid Project at the Robotics Club of IIT Kanpur, primarily focusing on its Computer Vision modules. In addition, I have undertaken numerous projects in diverse machine learning domains, including One-Shot Learning, Real-time Sentiment Analysis, and Detecting Semantically Similar Questions, among others. For a comprehensive overview of my projects, please refer to the “Projects” section on the page.
- Artificial Intelligence
- Human-Machine Interaction
- Robotics
-
B.Tech in Computer Science & Engineering, 2019
Indian Institute of Technology Kanpur (IITK)
Experience
- Working on identifying bottlenecks and enhancing the Android graphics rendering pipeline by adding improvements in the Android framework
- Developing an in-house Android profiling tool to benchmark Android rendering uniformly across various Samsung apps and devices
- Developed on‐device machine learning (ML) model that uses phone usage data to detect boredom while a user is using their phone
- Developed an end‐to‐end Android application to demonstrate the effectiveness of the boredom ML model to the stakeholders
- Worked on developing a deep neural network (DNN) model that uses phone usage data to predict gender and demographic age
- Developed the above model using TensorFlow Federated and Flower libraries to train it in a Federated Learning (FL) environment
- Developed a differential privacy-based ML solution for the problem of Privacy Protected Semantic Location Tagging
- Developed Android application for visualizing depth maps & 3D Point‐cloud from Time‐of‐Flight (ToF) camera feed in real‐time
- Developed gesture‐based UI features such as Zoom, Pan, and Rotation for the point‐cloud visualization module in the Android app
- Worked on developing an on‐device privacy‐preserving DNN model‐based solution for the problem of Next App Recommendation
- DNN model was designed under strict memory constraints to minimize network bandwidth costs during various FL execution steps
- Developed and trained the DNN model in Java using the DL4J library so that it can be trained and used on‐device on Android
- Developed an Android User Trial (UT) application that supported FL, model training, and inference on‐device for the DNN model
- Developed Neural Network (NN) model to predict the current location of a user based on their recent locations and time of the day
- Developed a simulation environment for replicating which cell tower in a given area a user would be connected to while in transit
- Developed an ML classification model to predict which cell tower a user is most likely connected to at any time of the day
- Top‐1 & Top‐3 prediction accuracies for the final model were 85‐90% and 90‐95% respectively on the evaluation dataset
- Implemented Convolutional Neural Network (CNN) models using Python and TensorFlow for an image classification problem
- Used Google ML‐Engine APIs to train various CNN models on the Google Cloud for accelerated experimentations and training
- Developed Server‐Client support using TensorFlow Serving for exposing REST APIs to generate predictions from the trained models
Research Work
Projects
