Aditya Gannavarapu

About me!

I'm a full-stack developer, machine learning engineer and designer, interested and passionate about human-computer interaction. I love the idea of 'AI with interfaces'. I'm confident about my creativity. I want to develop stuff that would make the world a better place, make people’s lives better. Understanding how humans process and solving real-life problems through this understanding is what drives me.

What I did so far (Click to open)

Web

Machine Learning

Game Development

Creative Space

Where to find me

Mutation Simulator

Drag and drop DNA to see how changes in the DNA sequence affect the resulting mRNA and amino acids.

Third Dimension

E-commerce furniture website built on Laravel where you can visualise furniture in 3D developed using three.js. To experience the visualization, Click Here

Gesture Pong, Gesture Controlled Pong Game

Using TensorFlow Object detection API, created a game which can be controlled using the gestures of your hand from video in front of webcam. Wave your hand in front of webcam to control the paddle. Inspired from the implementation of Victor Dibia.

Aranii Sansthan

NGO for young students to make their lives better.Initiative by our college professor.

MNIT (Modified)

Template of modified version of official website of MNIT,Jaipur.

Exploring ULMFiT for Genomic Sequence Data

This is an implementation of ULMFiT for genomics classification using Pytorch and Fastai. The model architecture used is based on the AWD-LSTM model, consisting of an embedding, three LSTM layers, and a final set of linear layers.

DNA sequences using Python

These notebooks explore FASTA, FASTQ and SAM, formats for storing DNA sequences. This is learned and implemented from JHU's Computational Genomics class.

Drawing with Machines (Sketch RNN)

Implemented using p5.js and sketch_rnn.js of Magenta, an open source research project exploring the role of machine learning as a tool in the creative process. Sketch RNN is trained on Quick Draw! doodle dataset of Google. It is generative model for vector drawings. If you start a drawing, it will finish it for you.

Speech Recognition (Wake-word Detection)

This is a demo that I’ve developed to recognize audio commands and trained a model on wake word detection, once it is activated, it takes user audio input and converts it into text to derive intents using NLP/NLU to provide feedback to the use.

InShorts News Sentiment Analysis

Web scraping InShorts news using BeautifulSoup and Requests. Performing pre-processing of text data using NLP libraries (NLTK and SpaCy). Sentiment Analysis was performed using AFINN Lexicon and TextBlob.

HipFlat Web Scraping

Pull data from Hipflat. This project will be using Selenium library to run the first get (request) to obtain all condo links, and then extract info elements using BeautifulSoup.

Image Style Transfer using TF.js

This is an implementaion of Neural Artistic Style Transfer running in browser using Tensorflow.js.Please follow the notebook to get an idea how this is implemented and the Github Repo to know more about running the algorithm in a browser.

Image Recognition (Heroku)

Image Recognition - Dogs vs Cats, deployed to Heroku using Flask backend. A six layer CNN bulit using tflearn using the Kaggle Dataset. All info regarding data and usage in Github repo.

Variational Auto Encoder

Implemented a VAE on MNIST dataset to generate new images. Encoding giving input noise with certain dimensions to latent space dimensions and Decoding the latent space dimensions to original dimensions.

Generating Music using LSTM

Using LSTM's in a Google Colab, we train a RNN with MIDI files of music to generate new music in MIDI format.

FIFA World Cup 2018 predictions

As a huge fan of football, using previous matches data avaialble from Kaggle, predicted the winners of each round ultimately predicting the winner of FIFA World Cup 2018.
PS : Predicted team didn't win! :P