Pega Data Scientist & Analytics

Prioritized Deliverables
1. Library of required queries/scripts to replicate the CDH customer contextual object in external systems (databricks/asl) for deeper analysis
2. Standardize format for executing key data retrieval steps for use by the broader team
a. Interaction to outcome attribution (account opens)
b. Model data to interaction mapping (model performance, predictor performance)
c. Member Profile to interaction mapping
3. Create notebooks for the broader team to use to answer specific questions
a. Distribution Analysis
b. Arbitration Analysis
c. Channel Engagement Analysis

Skillset:
The primary technical skills required would be familiarity with the databricks environment and proficiency with Python/PySpark and SQL. Pega CDH experience is preferred.
Some examples of the work as it directly relates to GEM
• Initial Analysis to Support New Model Related Features
o Propensity Thresholds
• Creating the back-testing approach (MDSA had no appetite at the time)
• Establishing baseline KPIs
• Creating the monitoring approach
o Initial Model Maturity Analysis (though Morgan's team is starting to be involved)
• Establishing baseline KPIs
• Gauging the impact of enabling the feature
• Creating the ongoing monitoring approach
• On-going Analysis
o Model Performance Monitoring
• Though MDSA owns the code to run the notebooks, when changes must be made to the code GEM is heavily involved in creating the new logic
o NBI Program Model Health
• This exists in some form today, but it is not in a state that is readily available to be shared with leaders in O&A, MDSA, or broader Marketing Broader O&A Analytical Gaps (Things Red, Sumant, and Tai typically scramble to create which should be readily available)
• "Actionable Monitoring Data:" Standardizing how we conduct this sort of analysis for consistency
o Capture when propensity scores are exceptionally low closer to real-time (1 day)
o Capture when actions are not providing value to their intended objective (acquisition, engagement)
• Eligible Audience Monitoring
o Identifying Members eligible for different actions/treatments (simulation environment can help after going live to a certain extent)
o Tying interactions back to key Member demographic data for more granular analysis (this sort of analysis should be standardized so it can easily be done
by all Members of O&A)

databricks
Python/PySpark
Pega CDH
SQL