Bobby Jacob Email & Phone Number

Chicago, Illinois, US

• Built several chat based demos for multiple use cases including:○ Document question and answer for product manuals○ Targeted marketing generated from model output of BigQueryML’s K-means clustering algorithm○ Product.
• Time Series Forecasting - Developed time series models for a major insurance company to predict claimsfrequency and amounts in two ways:○ Using BigQueryML ARIMA models○ Using custom Tensorflow modelsData used was.
• XGBoost model training time acceleration - UsedDask with GPUs for hyperparameter tuning on terabytesized datasets.
• Large scale signal-processing – Built predictive maintenance IOT models with data from thousands ofsensors using a Tensorflow autoencoder anomaly detection model.

Chicago, IL, US

• Installment Plan Modeling – Predicted customer default on installment plans for loss reserving and financing. Incorporated Experian credit reports as a source for the model.
• Fraud Modeling – Developed regularized logistic regression model to predict fraud using internal order characteristics and data from external fraud detection providers.
• Variant Testing – Ran A/B testing of website features to determine which performed best.

Chicago, IL, US

• Installment Plan Modeling – Predicted customer default on installment plans for loss reserving and financing. Incorporated Experian credit reports as a source for the model.
• Fraud Modeling – Developed regularized logistic regression model to predict fraud using internal order characteristics and data from external fraud detection providers.
• Variant Testing – Ran A/B testing of website features to determine which performed best.

Northbrook, IL, US

• Loss Modeling – Modeled loss using generalized linear models (GLMs) for auto, home and commercial products.
• Economic Modeling – Modeled bind rate for new business customers as well as renewal and midterm defection rates for existing customers using logistic regression in R and SAS.
• Credit Modeling – Worked on updating Allstate's insurance scoring model which predicted loss as a function of insured's credit history.