I have completed my B.E. in Computer Engineering from Thapar Institute of Engineering and Technology, Patiala, India.
Currently I'm working at MILVIK Technology Services, Bangalore, as a Software Engineer.
A Convolutional neural network model in Tensorflow for handwritten digit recognition using MNIST data set. Used two convolutional layers with max-pooling, two fully connected layers, Adam optimization algorithm and dropout for regularization achieving an accuracy of around 99.4%.
Neural style transfer
Python, Tensorflow, VGG-19, Transfer learning
Used VGG-19 model trained on ImageNet data to achieve transfer learning and calculated the content loss and style loss to create artistic style images.
Image caption generation
Python, CNN, LSTM, InceptionV3 model
Extracted features of images using pretrained InceptionV3 model, used CNN, LSTM (Long Short-Term Memory) and beam search to generate automatic descriptions for images. Model is trained on Flickr8K data set.
Neural machine translation
Python, Keras, Bi-LSTM, Attention mechanism
Developed a Neural Machine Translation model in Keras using Bi-directional LSTM and Attention mechanism to translate human readable dates into machine readable dates.
Object detection
Python, YOLO model, CNN, non-max suppression
Used YOLO ("you only look once") algorithm for object detection with non-max suppression to detect and locate cars in an Image.
Fake news classifier
Python, Naïve Bayes, SVM, scikit-learn, nltk
Used Naïve Bayes and SVM algorithms to classify news articles as fake/real. F1 score and accuracy of both were around 96% on test data. Implemented in python using nltk, sklearn and numpy modules.