• Developed an Explainable Hate Speech Detection System for Indic languages. Unlike typical sentiment analysis tools that use advanced models without explaining their decisions, our project focused on making the process transparent. We explored several techniques to understand how decisions were made, identifying that often, the focus was mistakenly on non-abusive words or punctuation. By employing a method called integrated gradients with the BERT model, we were able to show the importance of… Show more • Developed an Explainable Hate Speech Detection System for Indic languages. Unlike typical sentiment analysis tools that use advanced models without explaining their decisions, our project focused on making the process transparent. We explored several techniques to understand how decisions were made, identifying that often, the focus was mistakenly on non-abusive words or punctuation. By employing a method called integrated gradients with the BERT model, we were able to show the importance of each word in identifying hate speech. This approach improved our system's ability to detect hate speech and allowed us to create less harmful versions of the detected hate speech by masking aggressive words.• Designed a Two-Tower CNN Architecture for enhanced image recognition. Our unique approach combined a shallow network with a deeper ResNet-18 model, incorporating a bio-inspired layer at the beginning of the shallow tower. This innovative design significantly boosted the performance of deep learning networks by 5%-13% on popular datasets like CIFAR-10, CIFAR-100, and ImageNet-100. The success of this project highlights the potential of integrating biological concepts into artificial intelligence to improve technology.• Developed a Web Application for IC50 Prediction, making it easier to understand the potential effectiveness of drugs based on their properties and cell line information. The application uses three advanced deep learning models- tCNNs, Precily, and Paccmann, to predict IC50 values, facilitating faster and more accurate drug research. Show less

• During my internship, I authored Python scripts to analyze a third party’s customer purchase datafrom the past 3-4 years.• I also composed Python scripts to analyze the previous year’s sales data for an organic store.

During the internship, I worked on the e-commerce platform being developed by the company. Icontributed to following components of the platform:1. Payment Gateway Integration with PayTM2. SMS Gateway Integration with MSG91