- OneMicrosoftRecommendation Model : Developed the end-to-end model pipeline using Random Forest, boosting predictive lift and enhancing reason texts, which drove a 20% increase in sales.- AzurePropensityToBuy Model : Spearheaded the development of a Random Forest model to predict subscription propensity for 500K+ customers, achieving a 4X lift over the baseline model in Azure subscription predictions.- Model Drift Service : Designed and implemented a custom ML service on Azure to monitor model performance, detect data drift, and automate retraining for 50+ production models, reducing training time by 75%.- Model BackTesting Service : Developed a custom MLOps solution on Azure to evaluate model outputs and ensure alignment with business requirements, optimizing model performance for strategic objectives.

Developed a robust image recognition model for digit, alphabet, and icon classification across varyingresolutions, leveraging Convolutional Neural Networks (CNN) and TensorFlow Object Detection APIfor precise multi-scale detection.Fine-tuned multiple pretrained detection models, optimizing performance over 100K epochs, with SSDMobileNet achieving over 98% accuracy and superior training efficiency.

Developed long/short market neutral trading strategies in WorldQuant Fast Expression API topredict future stock returns. Involved in implementing various technical and fundamental indicators for Alphas to achieve optimalPnLs.