Click on each project title below to view github repository.

For this project, I built a supervised classification model which predicts which employee will stay with company and which employee will leave the company. This also estimates the probability of an employee leaving. The project also has presciptive solution where it shows what are the reasons behind employee atrition so an organization can take appropriate action to avoid employee attrition.

For this project, I used Facebook's Prophet package which predicts passenger Seats availability based on historic trends. I also created a flavour of algorithm where it predicts availability by different category in a loop. I added extra regressors for missing dates/data so that the model is not underfitting. These same extra regressors can be used for other variables, so the same code can be used for multivariate analysis. Plotly intercative charts are used for all forecasts so one can switch between different time periods in same chart without having to create multiple forecasts for multiple time periods.

This Project uses NLP libraries to analyze sentiments and tags them as positive , negative and neutral sentiments. It uses NLTK libraries to tokenize the words, and it also has word cloud to see what are most used words in a comment or conversation.

This projects aims to look at climate change by examining hurricane data from NOAA (National Oceanic and Atmospheric Administration) regarding the Atlantic basin. It does Geospatial analysis using python library folium to analyze hurricane tracks, landfalls and their impact over the years in US region. Folium maps used for this project shows the heatmap effect over the years theough sliders. This project can be used to understand geographical impact of any factor by analysing latitude and longitude data.

Inspired by paper https://arxiv.org/pdf/1506.06579.pdf. This project visualizes Neural Network Activation, Weights, Gradients to understand and interpret neural networks and what goes behind network activation and other aspects.

For this experiment, I built an unsupervised clustering algorithm which segments retail transactions in groups based on similarity and also detencts anamolous transactions.

This project contains analysis of Covid-19 cases worldwide and in US. It uses geospatial libraries to visualize data through interactive maps and it also uses Logistic regression sigmoid model to predict cases in future. It is an end to end machine learning solution.

My exposure to vision AI and tensorflow is limited, but in this project I attempted to create a binary classifier for Image and image segmentation.

This project uses shell and open source utility called "DATMO" to track tensoeflow models. Datmo is an open source production model management tool for data scientists