Barclays, Probability of Default, Calculated probability of Default for housing loan and credit card application using various modeling techniques. BharatPe, Time Series Anomaly Prediction, Predicted anomalies in the frequency of transactions on real-time data. Used Python and Prophet for the same. BharatPe, Classifying Transacting and Non-Transacting Merchants, Within 15 minutes of on-boarding, the time model could predict whether merchants would transact or not with 85% accuracy. This model brought down operations costs by 25%. Used SMOTE and Random Forest. BharatPe, Credit Risk Modelling, Built a credit risk model for disbursing loans based on various features such as Experian score, number of delinquencies, credit card, and so on. Used an ensemble of XGboost and Random Forest; scrapped contact information of customers from the undisclosed website for targeting new products to them. Adnet Global, Face Recognition, Built face recognition for identifying various Hollywood celebrities for UK-based fashion magazines. Used Python, OpenCV, dlib. The model could identify 90% of the celebrities correctly. Adnet Global, Time-series Forecasting, To fulfill the gap between demand and supply of human resources, time series forecasting was done to identify the demand for services. Projects were very seasonal, so SARIMAX and R were used which led to a reduction in cost by 40% PharmEasy, Prescription Recognition, Built an image classifier for identifying valid and invalid prescriptions. Used Python, OpenCV, and Resnet which reduced workforce cost by 30%. PharmEasy, Customer Segmentation, Clustered customers into 4 clusters based on usage patterns, this helped the Marketing Team to run targeted campaigns. PharmEasy, Database of Doctors, Built a database of doctors using Selenium and Beautiful Soup.