Python Data

PAHAL PATANGIA

*** **** *** ** ***********, MN **414 612-***-**** adc7ty@r.postjobfree.com

EDUCATION

UNIVERSITY OF MINNESOTA, Minneapolis, MN

Carlson School of Management

Candidate for Master of Science in Business Analytics May 2020 NATIONAL INSTITUTE OF TECHNOLOGY, Tiruchirappalli, India Bachelor of Technology in Electrical & Electronics Engineering May 2015

• Winner, South India Region - Barclays CIMA Global Business Strategy Challenge EXPERIENCE

CARLSON ANALYTICS LAB, Minneapolis, MN

Graduate Data Science Consultant June 2019 – Present

• Classified satellite images of farm fields to identify sustainable agricultural practices adopted by farmers using Mask RCNN framework in Pytorch for Land O’ Lakes

• Predicted Soybean prices using XGBoost, LSTM & TextBlob using tweets, USD index, S&P DCFI & temperatures date to recommend ‘sell’ positions resulting in potential savings of $5200 per stakeholder

• Revamped coupon mailing strategy by defining customer segments using K-Means clustering & Apriori association rule patterns leading to 25K potential customers in 6 months for a hospitality business in Minneapolis

• Developed an end to end real time fraud detection system on top of AWS Fraud Detector using SageMaker for modeling, Kinesis for streaming and SNS for live notifications

• Forecasted the number of calls up to 90 days, within 20% error using Facebook’s Prophet for optimization of resource allocation and eventually reduce manpower costs by $6M+ for Mall of America

• Built a classification model using LightGBM to predict loan defaults for Kaggle Home Default Credit Risk and explained the results using SHAP

• Presented a visualization application using Bokeh library in Python to demonstrate its business advantages as compared to Tableau and Plotly for Carlson School Trends Marketplace contest FAIR ISAAC CORPORATION (FICO), Bangalore, India

Analytic Science Consulting - Consultant II June 2017 – June 2019

• Built a customer credit default prediction ensemble model (Product level to Customer Level) using Generalized Additive Models (GAM) to preemptively detect customer's riskiness arising from any of its accounts

• Developed an underwriting model using Gradient Boosted Decision Trees (GBDT) leveraging credit bureau, transaction, and demographic data, thus increasing the bank's acquisitions by 14%

• Optimized origination risk-return tradeoff for Small Business Loans by evaluating segmentation schemes & building models using Segmented Ensemble Modeling (SEM) technique & GAM respectively

• Designed and developed IFRS9 compliance models (PD, EAD & LGD) for a Latin American bank leveraging Markov Chain and Decision Tree algorithms to maintain incremental expected credit loss within 17% of threshold

• Worked on building revenue and attrition models for a bank using Gradient Boosted Regression Trees and logistic regression respectively

Analytic Science Consulting – Consultant I July 2015 – May 2017

• Developed an application fraud model using XGBoost algorithm to eventually increase detection rates by 32%

• Validated & set up tracking processes for traditional risk solutions by observing statistics like Population Stability Index (PSI), Characteristic Stability Index (CSI) and model's KS, Gini & Divergence performance metrics

• Streamlined prototype data cleaning processes in Python to accelerate daily tasks for a 56-member analytics team SKILLS

• Technologies: Python (numpy, pandas, sklearn), SQL, R, Scala, Hadoop, Hive, Spark, AWS, Tableau

• Analytics: Machine Learning, Neural Networks, Big Data, Databases, Visualization, A/B Testing, Time Series

Contact this candidate