Projects

Worked with Anheuser-Busch InBev (AB InBev) on developing clustering models to perform customer segmentation. Developed a full-stack application for the model using Poetry with Flask, PostgreSQL, and Angular. Containerized the application using Docker.

I developed an app to help you answer how likely (or rather, unlikely) natural or artificial disasters are in an area. The app utilizes open datasets from the Federal Emergency Management Agency to provide historical data on the number of disasters in your state and the average economic costs associated with those disasters.

Developed an emotion recognition model on the Emotional Prosody Speech and Transcripts dataset using MFCCs and CNNs and achieved an accuracy of ∼80%.

Kervolution-based SubSpectralNet model using Gaussian Kernels for Acoustic Scene Classification, achieving ∼75.76% accuracy on the DCASE 2018 dataset. Replaced the linear convolutional operation with the non-linear kervolution operation via kernel trick for increased non-linearity.

Link to paper: Acoustic Scene Classification Using Kervolution-Based SubSpectralNet

Extracted real-time stock data and news using Flask and Apache Spark to load the data into MongoDB. Combined this with historical user data to generate recommendations. Orchestrated the ETL pipeline with Apache Airflow.

Created a full stack website with detailed booking options for users and CRUD functionality for employees. Designed the MySQL database schema with 2NF normalization and security measures against SQL injections. Designed the front-end using AngularJS and created Flask endpoints to connect to the database, providing encryption for sensitive data.