REST API to convert Kannada sentences to English using Seq2Seq Nueral Network model. Implemented attention mechanism. Deployed the transalation model on Google Cloud using cloud functions. The transaltion model is implemented using PyTorch.
Web application for product recommendations to customers based on their reviews at the online grocery store. The web application is a prototype to mimic a real time application where the recommendations are rendered with product images for a given input string as a review.
Implementation of Spark context, Spark SQL context on Amazon Tweets data set with 400k Tweets. Analyzed the tweets on the busiest day to find the words that were repeated the most in the selected tweets.
CNN Encoder - RNN(LSTM & GRU) Decoder model for captioning an image with visual attention mechanism. Building encoder-decoder models with global and local attention mechanism
Finding the contextual similarity between documents using Jacard and Cosine similarity metric. Implemented basics of text similarity on multiple files and presented the analysis.
Analyzed web analytics data of 500+ customers with 700+ variables to improve the conversion rate of visitors to customers. Presented XGBoost model with suggestions to improve it's performance.
Identified factors causing chronic kidney disease. Built the predictive model to identify subjects, who could be potentially affected by CKD. Interpretation of the statistical model to estimate the impact of each factor leading to CKD.
Built the Predictive Model to estimate the price of treatment given the clinical factors at the time of admission. Feature engineering of clinical factors to identify portential predictors for the price estimate.
The Machine learning based research study aims at examining and monitoring the various depression markers by analyzing the tweets of self-declared depression patients on Twitter. Temporal analysis of pre and post diagnosis of depressed individuals.