Hello! I'm

Prateek Agrawal

I'm 21

I Data and Dogs

I'm a Machine Learning Enthusiast

Trainer of Models and Networks

About

Hi! I'm Prateek Agrawal

I am a third year undergrad currently pursuing my B.Tech and M.Tech Dual Degree in Computer Science at Indian Institute of Information Technology Design and Manufacturing, Kancheepuram, INDIA.

I am a Machine Learning enthusiast looking to new challenges everyday. I also like to learn and share my knowledge and is a part of various communities to explore to various facets of the field.

" I LOVE DATA and DOGS   "

Checkout my resume

Contact me here!

 Email: prateekagrawaliiit@gmail.com

 Phone: +91 - 9999211405

What I do

My Interests

Machine Learning

Recommender Systems

Data Science

Web Applications

NLP

Computer Vision

My Hobbies

Gaming

Netflix

Reading

Writing Notebooks

Competitive Programming

Travelling

Portfolio

Checkout a few of my works

Web Application

We know the worth of something when we have lost it.

LoFo is a web application developed to tackle the day to day problems of Lost & Found by the residents of Indian Institute of Information Technology Design and Manufacturing,Kancheepuram,India.

LoFo has been developed on CodeIgniter Framework by using technologies like HTML,CSS,JS,PhP and MySQL

View Project

Natural Language Processing

Sarcasm Really ?

  • Built a Sarcasm classifier using Neural Networks and word embeddings on Rishab Mishra's 'News headlines sarcasm dataset '   and achieved an accuracy of 99.98% on training and 88.6% on test data.
  • Technologies Used : Keras, Tensorflow, Numpy, Pandas

   View Project 

Computer Vision

The ridiculous Not Hotdog app from 'Silicon Valley' is real

  • Built an Image classifier that classifies the image as whether the image is of a hot-dog or not. Inspired from the famous show 'Silicon Valley'  the model was a good opportunity to learn new things.
  • Used Transfer learning to train the model on ResNet50 and VGG16 models and achieved an accuracy of 84.7% for ResNet50 model and 83.9% for VGG16 model.
  • Technologies Used : Keras, Tensorflow, Numpy, Pandas, ResNet50, VGG16

   View Project 

Recommender Systems

Market Basket Analysis

  • Performed Market Basked Analysis on a dataset of transactions for a retail store in France and one in Germany to obtain patterns in the sale of products.
  • Extracted patterns using Association Rule Mining by Apriori and FP-Growth Algorithms after preprocessing the data into the desired format.
  • Published a notebook on Kaggle on step to step guide to perform Market Basket Analysis using Python.
  • Technologies Used : Python, Numpy, Pandas, mlxtend.

 View Project 

Natural Language Processing

IMDB Movie Reviews Classification

  • Experimented with GRUs, Single layered LSTM , Multiple layered LSTM and Convolutional Neural Networks to perform sentiment analysis and predict the label for IMDB Movie reviews and obtained best results with LSTM based model with an accuracy of 85% on test data.
  • Used plain text as well as sub words to find out which model is best suited for the classification.
  • Used tokenizers and vectorizers to process the text data to be fed into the neural network.
  • Technologies Used : Keras, Tensorflow,LSTMs, Gated Recurrent Unit, Recurrent Neural Network, Convolutional Neural Network,Python

 View Project 

 View Embeddings Projector 

(Please use Spherize data for better visualisation)

Web Development

Vidhai

VIDHAI is a non-profit initiative established by students of IIITDM Kancheepuram thereafter endorsed by the administration. Currently, about 76 students are volunteering for VIDHAI.

Built the website using plain HTML,CSS and JS

View Project

Computer Vision and Transfer Learning

Cats Vs Dogs

  • Achived an accuracy of 99.7% using InceptionV3 transfer model, Convolutional Neural Network and Deep Neural Network.
  • Preprocessed the data using Image Generators and used data augmentation to capture more features that the network wasn't able to .
  • Experimented with various other Convolution Neural Networks with and without data augmentation and achieved an accuracy of roughly 73% from them.

 View Project 

Computer Vision

SIGN LANGUAGE MNIST

  • Achived an accuracy of about 95% using Convolutional Neural Network and Deep Neural Network to classify hand signs into numbers from (0-9) instead of the traditional MNIST dataset.
  • Experimented with various methods such as data cleaning, data processing , data augmentation and hyperparameter turing to improve the performance of the model.
  • Technologies Used : Keras, Tensorflow, Numpy, Pandas and Python.

View Project

Portfolio

I love to share my achievements

Projects Done 0
Cups of Chai 0
Hours in CSGO 0