Pragyesh Mishra Email & Phone Number

Noida, UP, IN

Foster City, California, US

San Jose, California, US

• ❖ Client- US Based RetailObjective: To forecast spoilage of Pizza subcat in 2500 stores at a Store*Subcat*Hour level.My role was-
• To develop a forecasting model that should be able to predict sales forecasting at an hourly level and generate a Microstrategy report.
• To calculate WAPE numbers at a store level and also predicting the under-prediction andover-prediction w.r.t spoilage.
• Use MLFLOW to track and record model prediction WAPE of the Holt-Winters technique of forecasting which gave us the best results.❖ Client- US Based ECommerce➢ Databricks Migration(MLOPS)-Objective: The engagement.
• To setup ML FLOW Databricks for ML Model creation, training, tracking runs andexperiments for hyper parameter tuning and deployment.
• Manual implementation of Databricks pipeline to deploy it as Kubernetes pods bycontainerization using Docker and hitting an End point API using Flask. Thenautomating the process as CI-CD by using Terraform scripts and.

Delhi, Delhi, IN

• ❖ Client–AustralianSkills - Feature Engineering, Feature Selection, CNN, PCA, EDA, Classification, LinearRegression, Logistic Regression, Pandas, NumPy, Scikit-learn, Seaborn, Matplotlib DecisionTree, KNN, K-Means.
• Defined Churn definition with customer (Label identification).
• Explored and analysed the customer profiles (EDA) – Perform correlational analysis to identify features that have maximum impact on a MSISDN churning.
• Trained a classification model for predicting the propensity of churning in future.
• Validation- Performed cross validation to confirm model’s performance on unseen/new data.

Irvine, California, US