Passionate, pragmatic problem solver and programmer, proficient in Python programming, pursuing simpler solutions to complex problems. I am a CS graduate student at ASU. I love building android apps, creating deep learning models to solve real world problems (Time series, language processing etc) and solving coding challenges. I excel at software development using python, Java and C++ and have 2+ years of experience at Deloitte doing the same. I have interned as a data science intern at Ziggurat during June 2020 and volunteered as a full stack developer at Corona Connects during May 2020. This portfolio is intended as a showcase of my projects and my technical skills. And if you get bored, you can play my version of the classic Atari game Asteroid Blaster.
A 19 layer VGG convolutional neural network that recreates images in the style of classical paintings. Below are the original (lef) and model generated (right) images of a bridge captured at night time using my mobile phone. The image on the right is rendered in the style of Van Gogh's famous painting The Starry night.
Reduced bias in recidivism prediction data for analyzing criminal’s risk of re-offending and increased fairness index from 44% to 78%. Model was trained to analyze a the COMPAS recidivism predictions of more than 10,000 criminal defendants in Broward County, FLorida and compared to actual recidivism rates over a period of two years.
Created a machine learning model using KNN technique to predict handwritten digits with more than 96% accuracy. MNIST dataset used for this project contained over 70,000 64x64 images of handwritten digits. Predictions were made by calculating Euclidean distances to find K nearest neighbors and using maximum voting strategy.
C++; Python (Tensorflow, Keras, Numpy, Pandas, Scikit-learn, Matplotlib);
MATLAB; Java; Javascript; SQL; BigQuery; CSS; HTML;
AWS; Docker; Google cloud (GCP); Eclipse; Visual Studio; Conda; Jupyter; Axure; Microsoft Suite; Adobe Suite (After Effects, Premier Pro, Audition); Jenkins; Jira; Dynatrace; Latex;
Linux distros (Ubuntu, CentOS, RHEL); MacOS; Windows;
(Fourth course in deep learning specialization by deeplearning.ai)
Building a convolutional neural network, including the recent variations residual networks (resnets) and Alexnets.
Applying CNN to image, video, 2D and 3D data.
(Third course in deep learning specialization by deeplearning.ai)
Understand how to - diagnose errors in machine learning systems, prioritize most promising directions for reducing cost function and understand complex ML settings to supass human-level performance.
Applying CNN to image, video, 2D and 3D data.
(Second course in deep learning specialization by deeplearning.ai)
Understanding and implementing neural network techniques, including initialization, L2 and dropout regularization, Batch normalization, gradient checking.
Implementing and applying a variety of optimization algorithms, such as mini-batch gradient descent, Momentum, RMSprop, Adam optimization and convergence checking.
(First course in deep learning specialization by deeplearning.ai)
Building, training and applying fully connected deep neural networks.
Implementing vectorized neiral networks
(Second course in Algorithms specialization by Stanford)
(First course in Algorithms specialization by Stanford)
(Implementing advanced Tensorflow models on Google Cloud Platform (GCP) including techniques such as pipelining, one-hot encoding.)
(Human research - Social & Behavioral Research)