About Me

I am a Machine Learning Engineer (Research) currently working with Advanced Tech Labs of Samsung Research Institute, Bangalore. I did my B.Tech. from Indian Institute of Technology, Guwahati with a Bachelors' thesis in Computer Vision and a minor in Product Design.

I believe myself to be a strong-willed, responsible and pragmatic person. To feed my curiosity with the abilites of computers, I always thrived to expose myself to a myriad of paradigms that together constitute computer science. I love to travel and capture moments with my camera.

If you wish to know more about me, download my Resume.


Photo-Realistic Emoticon Generation Using Multi-Modal Input

International Conference on Intelligent User Interfaces (ACM IUI) 2020

We trained a Generative Adversarial Network (Pix2Pix like U-Net) that generates photo-realistic emoticons from hand drawn sketches. This enables the user to personalize the emoticons she/he wishes to use. We also enabled users to write keywords to further refine the generated images. Sketch images as input are translated to emoticon images with high levels of personalization. It is accepted as a short paper. For further details please visit here

Image Memorability: The role of Motion and Depth

International Conference on Image Processing (ICIP) 2018

We analyzed the role of motion and depth cues in determining how memorable or forgettable (using Image Memorability) a particular image is. We use an ensemble based deep neural network to predict the Image Memorability scores using deep object, motion and depth features and achieve a rank correlation of 0.65 on LaMem dataset which is very close to human consistency. For further understanding please click here

Misbehavior Detection in C-ITS Using Deep Learning Approach

International Conference on Intelligent Systems Design and Applications (ISDA) 2018

This work revoles around the use of data driven Artificial Intelligence to predict traffic node misbehavior in Cooperative - Intelligent Transportation Systems (C-ITS). A misbehaving node can feed malicious information like wrong traffic congestion details to centralized server. We use Deep Learning to predict anomalies by measuring deviation from expected (predicted) traffic congestion readings.


Samsung Research Institute Bangalore

Summer of 2017 were spent working in Samsung's Knox Division for a project on classification of Document Images based on its confidentiality using Deep Neural Networks.

Use of Convolution and Recurrent Neural Nets for classification followed by an android app to click document pictures, predict confidentiality and move them to secure space if needed.

Web Developer for GT-Tech

Created a Django based online merchant platform for a Bussiness to Bussiness startup called GT-Tech. The startup aims to mitigate the effects of middlemen in small scale cottage industry of Assam by providing an online tool that joins raw material buyers to sellers and vice versa.

Android Developer

Summer of 2016 were spent building an Android Application for a startup. The Google Maps based application was tasked to get user preferances, details etc. and provide suitable outcomes on maps with shortest routes and turn by turn navigation features. Volley based server-client request handling, Firebase based Social login and feedback collection for specified users were implemented.



2014 - 2018

Computer Science and Engineering with the Indian Institute of Technology Guwahati.

CPI: 8.60/10.0 (Without Minor)

Senior Secondary - CBSE


Modern Senior Secondary School, Kota.

12th Boards Percentage: 93.8%

Secondary - CBSE


Modern Senior Secondary School, Kota.

CGPA: 10/10


My bachelors degree has exposed me to a wide variety of technologies thereby diversifying my knowledge, never making me take a parochial approach towards solving problems.

Programming Languages


Principle choice of language while competitive coding. C language was also used of assignments on pintos, ns3: network simulatior.


Python programming language was used for a number of applications like web crawler and search engine, video - recommender, Raspberry Pi2 server for Iot based project and Deep architecture implementations using Tensorflow.


Mainly used for Android application development, along with a client-server multi-threaded application as part of Programming Languages Lab.


Used with Visual Studio mainly for the development of interactive system for ONGC's oil rig decision support system, as part of sophomore year project.

Web Technologies


Leveraged the ease of web-app making with python-django framework to create website for a B-2-B startup. Created an end-to-end merchant platform for online shopping.


Used the light-weight python-flask for back-end of a video recommender application that involved efficient deployment of three Database Management Systems.


Used to render front-end of sophomore year projects on web search engine, Assembler Linker Loader etc.


Building of interactive front-end for all web applications mainly using the Bootstrap framework.



Proficient in use of tensorflow library for creating deep neural archtecture. With various ongoiong projects in Samsung, Bachelor Thesis project, summer research Intern and a few course projects on Deep Learning, achieved considerable proficiancy in use of tensorflow.


Used the library with python for a number of sub-tasks like image resizing, noise reduction and other preprocessing tasks.

Android Studio

Developed an android application as part of sophomore year internship for a startup. Also created an android app for image classifier using tensorflow as part of pre-final year Summer Internship with Samsung.

Visual Studio

Used the tool as part of Sophomore year project for ONGC's oil rig decision support system. And also for a multi-player hide and seek game as part of Code-Fun-Do.

Database Management


MySql was used at multiple places as part of my Bachelors program, most prominant ones being design of schema for Django website and for video recommend system where user logs and history were stored in mysql.


Use of document based noSQL database for storing video data as JSON files.


To store weighted relationships between a number of videos based on multitude of possibilities. Such relationships further play a major role in nect video suggestion.