1. Identified improvements in the algorithm to reduce the false positive alerts by 15%.2. Worked on building an ensemble/ best of time series forecasting models (using Prophet, ARIMA,XGBoost, linear regression) to identify unusual trends as well as anomalies.

1. Analyzing sales patterns for over 600 kiosk locations and 1M monthly transactions to understanddemand transfer in different locations and make optimal assortment decisions at each kiosk.2. Built similar product recommendations by utilizing sales patterns and product descriptions, whichhelped in improving product placements and increasing sales.3. Employed multi-variate time series data and local factors to predict sales, which helped in improvinginventory management. Built a python dashboard with actionable insights for business which helpedimprove stockouts by 10%.4. Trade-off between compute time complexity, scaling and algorithm performance. Family of algo-rithms considered - Similar items, Complementary algorithm, Cold Start Algorithms, User RestartAlgorithms, Repurchase and Next Basket algorithms. Studied deployment of embedding basedalgorithms such as word2vec, BERT, nmslib, FAISS .5. Implemented impact of various techniques for customer loyalty analysis such as lifetime value, RFManalysis, cohort retention on recommendation metrics.