Built and deployed custom DNN architectures based on Deep Cross Networks.

Led the development of batch processing techniques in TensorFlow, optimizing deep neural network (DNN) inference costs and driving $2 million in annual cost savings for the organization.

Improved DNN model performance using clustering techniques like Glove, and asymmetric loss functions.

Executed A/B tests with significant business impact, achieving a 5% uplift in user acquisition

Spearheaded DNN model deployment, ensuring rigorous validation and fostering interdisciplinary collaboration for seamless integration into production.

Recruited and mentored high-performing data science teams across global operations.

Built data processing pipelines on AWS using Python to streamline training and inference of models.

Built An aspect-based Sentiment Analysis engine to extract insights from customer review data. Received appreciation from client and Fractal.ai.

Designed and developed a text classification model to save $30k and 5000 man-hours using Logistic Regression, SVM, TFIDF, n-gram & cosine similarity. This model categorizes products based on descriptions.

Coached and mentored 4 data science interns; 3 of them were later hired into full-time roles.

Deployed a machine learning model on AndroidOS to detect human activity recognition based on sensor data.

Increased the efficiency of weekly sales report generation by 75%, using Python and SQLite for automation.

Improved accuracy of stock price prediction model using text (news) data by 5% using Convolutional Neural Network (CNN), Information Extraction (IE), and Neural Tensor Model.