Ananta Arora

Built predictive models to identify physicians who can potentially prescribe and will continue to prescribe a novel drug in the US Market. Built end-to-end Machine Learning pipeline using MLflow from myriad sources on data bricks with in-depth understanding of data sources, disease understanding & business rules. The pipeline included results from algorithms like Decision trees, XGBoost, Gradient Boosting Neural Networks, Ensemble and Random Forest Classifier to choose the best method to identify potential prescribers. Worked on Large Language Models (LLMs) to generate personalized emails and content for physicians to prescribe newly launched drug in the market. Monitored production model performance and retrained models. Delivered physicians recommendations along with optimum detailing, emailing frequency and enabled monthly run & tracking. Honoured to have been awarded the prestigious Bravo accolade in recognition of my outstanding achievements and unparalleled delivery of top-tier predictive models.

Risk of Skin Cancer Detection (Melanoma and Non-Melanoma) with a Machine Learning approach over image data from ISIC and PAD-UFES-20. Feature Extraction on the basis on color (mean, standard-deviation, skewness, and Kurtosis) and texture-based features using Wavelet Transform Coefficient for each rgb component. Detection using Support Vector Machine (~75%) and Gradient Boost Algorithm (~81%). Built extensive visualizations on 60+ million obit data stored in MySQL database for real-time insights. Automated the process by integrating Square’s API with Python and MySQL database for 8+ million restaurant data for real time insights of customers and sales summary. Closely worked with business development and marketing team in US and helped them with their requirements. Led the team for SSA Extraction and Cleansing data rules. The Obituaries’ data was extracted and scraped from funeral home websites using Scrapy. Collected around 90% of US obit data and automated data extraction, cleansing, standardization, and consolidation techniques. Built extensive visualizations using Power Bi on 60+ million obit data stored in MySQL database for real-time insights. Closely worked with business development and marketing team in US and helped them with their requirements.

Optimized Search Engine using TensorFlow Transformer model to optimize search query results for customer satisfaction. Improved the performance of the model from 63% to 76% using dataset balancing and hyperparameter tuning techniques.

Worked on research of latest technologies used in the field of AI. Built ensemble model to predict whether a candidate will register for any online course offered by one of the leading online academic institution-INSAID and made a flask app using RESTful API.