Andres Arcila, Ph.D.

Experience

• 

  University of Waterloo

  Apr 2019 - now

  Sessional Lecturer

  I teach/have taught a variety of courses in the Department of Economics. ECON101, ECON201, ECON322, PS624, PS626, PS629.

• 

  TD

  Jul 2020 - Oct 2021

  Credit Risk Model Associate

  I worked as an Associate at the Quantitative Associate Rotational Program at TD bank. This program is designed to provide associates with comprehensive experience in the different quantitative areas within the Risk Management department. -Basel III credit parameters (PD, LGD, UGD), regulatory & economic capital for AIRB credit risk, credit allowances, and Risk Ratings.-Methodology development related to credit risk and operational risk.-Canadian IFRS 9 and stress testing team on projects related to recent OSFI MST. - Validation and approval of analytical models used for measuring the risk exposure inherent in retail portfolio, as well as, models used for risk capital calculation of Retail products. Show less

• 

  Anheuser-Busch InBev

  Oct 2021 - Aug 2023

  Senior Research Data Scientist

  As a member of the Data Science team, I collaborate closely with demand model estimation, forecasting, and promo optimization groups to design and conduct experiments aimed at identifying causal effects resulting from their recommendations. My responsibilities include:• Developing a Bayesian Demand model and ML optimization routine to obtain price elasticity and contributing to the development and creation of the price optimization engine for price increase decisions at different geographic levels. This resulted in driving over 1 billion USD in revenue for the company.• Conducting Randomized Control Trials, A/B tests, and utilizing observational methodologies, such as Matching Estimators, Difference-in-Difference, and Instrumental Variables, to estimate the average treatment effects of our recommendations for the North American Market. • Participated in the discussion and implementation of a recommendation system using a demand for heterogeneous product models. This system provides stores with a curated selection of candidate products, strategically chosen to optimize their revenue generation.• Experience moving ML models into production. I have successfully led projects from data pipeline development to model deployment, ensuring seamless integration into the production environment.• Creating Time series and machine learning methodologies for forecasting.• Extensive experience with Big Data Framework like Pyspark and Hadoop and analytical software packages like R, Python, SQL, and Version control (Github, Bitbucket). • Experience in software engineering-level coding and adept at deploying machine learning models to production using MLops frameworks. Show less

• 

  Scotiabank

  Nov 2023 - May 2024

  Senior Manager, Credit Risk Model Development

  In my role as Senior Manager of Model Development, I oversee projects dedicated to the creation and maintenance of Statistical and Machine Learning Models for Risk Assessment. Our primary focus is on estimating Risk Weighted Assets (RWAs) for the Retail Models team. These models play a crucial role in evaluating the risk associated with a portfolio valued at approximately CAD $5 billion within the Canadian Retail sector.My responsibilities encompass the application and management of a comprehensive model monitoring framework, which includes Effective Model Management (EMMs) and Model Monitoring Plans (MMPs) for Canadian portfolios. This framework extends to cover RWAs and Expected Credit Loss (ECL) models, as well as Scorecard models, which incorporate advanced Machine Learning techniques to enhance their predictive power. Show less

• 

  Affirm

  May 2024 - now

  Senior Data Scientist, Growth Analytics

  My work at Affirm focuses on applying observational methods for causal inference within the growth analytics team. This involves leveraging data from natural experiments and observational studies to identify and measure the impact of various growth strategies. - Utilize advanced techniques: Synthetic Difference in Differences, Matching, Meta Learners, and Contextual Multi-Arm Bandits.- Developed meta-learner and uplifting models to identify high-incrementality customer segments, leading to a reduction in marketing spend.- Employed synthetic control methods to evaluate a new product feature, resulting in a better understanding of their incrementality. - Bridge the gap between data science and business strategy to optimize growth initiatives. Show less