Hi, I'm Prakhar.

Data Scientist

About me

Who I am

Hello, my name is Prakhar Thapak. I am a data scientist based in Bangalore, India. Since the age of 15, I have delved into numerous programming languages. My interest lies in deep learning and natural language processing.

What I do

I solve complex business problems by making use of state of the art machine learning techniques. I also like developing engaging user interfaces . I find the process of shaping an online identity through engaging user interaction fascinating. Since an image is worth a thousand words, scroll down to my portfolio for a few thousands words worth of images.

Neat facts

Coding: Apart from Python,C++ and JS at times I've worked with Ruby and Node.js.
Cryptography:I am very much interested in cryptography and have used elliptic curve cryptography as well as various techniques for my project Visual color cryptography using elliptic curves
Hobbies: I play chess. I'm a casual gamer, huge GTA and Assassin creed fan.
Languages: I speak and write English,Hindi and decent in French!

My tools

Python

95%

Tensorflow

85%

Pytorch

80%

C++

80%

Dash

95%

MongoDB

82%

Django

90%

Flask

95%

Portfolio

Recommendation system

Built a recommendation system using recurrent neural network (bilstm cell) and convolutional neural network (inception v3).

UI development with dash and flask

Used dash and flask to develop a UI for visualizing machine learning models (open source)

GitHub

Object Detection

Worked on Object detection using Faster R-CNN as well as Yolo v2 being the underlying architectures.

Self Attention GAN

Worked on SAGAN that was used to improve the image quality as well as used to generate words in the caption from image by attending a specific part of image .

Neural Machine Translation

Worked on Neural Machine Translation using seq2seq model with attention

Visual color cryptography using Elliptic curves

The purpose of this project is to first implement Color visual cryptography using substitution method which takes the plain text and converts it into ciphertext by using ASCII values of the character and the key which is generated using the color channel selected and the color block size and then implement ECC which works on the principle that finding the discrete logarithm of a random elliptic curve element with respect to a publicly known base point is infeasible which is termed as ECCDLP.