To minimize any future outages and increase the productivity of developers of a company CCC IS, a prediction software was designed & developed using LSTM model of Tensor-flow. The objective was to predict error for each server for next coming 7 days on a daily basis. This was the capstone data practicum project. The error pattern was analyzed and detected from the real-time dataset.

My contribution was to clean and extract data to put into a MySql database. I developed the Python code to implement the LSTM model.

To make computer more helpful a knowledge-based AI agent to mimic human skills to solve Raven’s Progressive Matrices using a Pillow image processing library of Python by reading pixels from the given images.

I was the sole developer.

To improve diagnosis process and hence reduce doctor’s time to diagnose and help to treat more patients in less timing frame, the real-time clinical dataset provided by Illinois Institute of Optometry was analysed and classified all patients record to expedite or differ the treatment of more or less serious patients respectively in order to suggest doctors while making decision using decision tree, logistic regression and random forest classifier

My contribution was to read the dataset from pickle repository and build machine learning models (Naive Bayes, Random Forest and Decision Tree)

To increase revenue of online movie providers or DVD sellers, improved movie recommendation system by making it more reliable in terms of accuracy, 1.5 million movie records have been crawled from IMDB to generate training set for 28 target genres and then applied Cosine similarity algorithm using TF-IDF Vectors to classify the relevant genre of new movies as test input.

My role was to web scrape data from IMDB and build the model using TF-IDF.

To improve the performance of OCR used to detect ZIPCODE during Postal Services, a digit recognizer system based on machine learning techniques was designed & developed. Training data was taken from MNIST to build and test the model using neural network and random forest algorithms to classify digits from 0 to 9.

My role was to build a random forest model.

Designed & developed a DBMS to store and retrieve records from secondary memory by building storage manager, buffer manager, and record manager using C

My contribution was to develop the code in C.

Designed & developed all assignments using R under the course CSP571

To make electricity bill more economic for American consumers and hence to prevent Global Warming to some extent, an optimization algorithm using a day ahead hourly pricing data from ComEd (an electricity provider in the USA) was implemented to produce a most economic schedule for each home appliance to operate within specified time period across 24 hours of the day

I was the sole developer.

Designed & developed an image search engine for the Scarlet Hackathon competition.

I was the sole developer.