• Led sanitized data layer buildout for US Telecom for near real-time customer journey analysis. Processed >1TB data/day using PySpark, BigQuery, DBT, and Airflow. Rewrote data science POCs, reducing model runtime by ~96%.• Built reporting for a promo optimization tool at a global CPG manufacturer. Completely replaced prior underlying Databricks data model and DAX measures to improve rendering times by 60-90%. Added logging to catch upstream data issues. Created C# deployment tool automatically deploy to >30 different markets, as well as automatically generating partitions and RLS to improve performance and increase security flexibility. • Built tools in PySpark to quickly generate GBs of complex, realistic synthetic data. These datasets supported Digital Twin visualization demos at clients, speeding up visualization feedback.

• Migrated healthcare data from Lotus Notes to Salesforce using SSIS and SQL Server. Worked with Salesforce team and client to develop a denormalized schema for the information. Handled corrupted input files, fuzzy deduplications, and incoherent data elements. Built out automated testing, input validation, and a multithreaded file uploader in C#, which was ~15x faster than the SSIS implementation.• Taught an “Intro to Python” course to Associate Consultants, going over basics of data analysis and data visualization. Ran weekly TA sessions focused on coding design considerations. Additionally, introduced formatting, testing, and typing concepts to students.

• Automated AWS deployments using CloudFormation and CodePipelines. All core projects now generate at least a dev and prod environment, allowing much quicker and less painful deployments of upgrades.• Rebuilt AWS security to significantly reduce limit unnecessary access to AWS resources using Network ACL and security groups. Projects now default to using their own unique security group allowing access only to other resources in the security group and access from computers within the CTA’s network. Set up monitoring to ensure adherence to these rules.• Orchestrated all AWS ETLs through Airflow, using custom operators. Operators include wrappers around Lambda, Glue, EMR, and ECS, as well as higher-level operators controlling extraction from database types to S3 and from S3 to AWS. This allowed much more fine-tuned control over simultaneous connections to production databases and had much better monitoring. Created script to automatically tag resources within a DAG, increasing visibility into how money is being spent. Exposed the Airflow API and console to analytics departments, making the ETL process much more transparent.• Having tagged resources either through Airflow or CloudFormation, created report that would identify old dev environments and orphaned resources. Deletion of those resources reduced costs by ~12%.• Optimized Redshift instance by running “analyze compression” and following suggestions as well as making use of sort- and distkeys more frequently. This resulted in substantially increased performance for a few of our largest tables.• Created common resources such lambda layers, a base lambda docker image, and an AWS utility package, substantially decreasing the amount of duplicate and near-duplicate code being deployed. Also replaced many per-table lambda and glue scripts with individual, parametrizable scripts where appropriate.

• Led the infrastructure rebuild of the Performance Management department. Created and integrated tools on the department server for automating emails, simplifying my AWS Redshift uploads, managing credentials, email tracking, git, CI, Docker, and a Wiki for resource sharing. I also created monitoring scripts to quickly warn when dangerous activity occurs. • Revolutionized the department’s main workflow’s to <5% reliance on Microsoft Excel. Working with IT I gained access to new data sources to avoid using canned reports and taught other analysts how to use libraries like Pandas and Openpyxl to replace excel macros.• Built and optimized daily data extraction from every department including emails, Excels, PowerPoints, PDFs, websites, databases, and custom applications. This allowed a 40-page PowerBI report presenting two years of data for each metric to be created and vetted in ~45 man-minutes every morning.• Created a model that predicted the top 10 individuals most likely to have an “unexpected” absence. Averaged 70% accuracy.• I created both a tool to automatically generate the daily train schedule by automatically parsing PDFs and Word documents. In addition, I created a calendar displaying the service changes for each line. Previously bulletins were emailed and manually transcribed, resulting in reduced service accuracy. • Expanded my reporting to include many interactive web tools, allowing individuals to see trends and metrics unavailable in the standard 3rd party reports. One of the results has been a decrease in the amount of overdue work since the roll out of these tools.• Improved department culture, supporting the younger analysts and creating a stong team spirit. This included fortnightly lunch outings, emphasizing sharing credit in department meetings, and being a resource for coding help.

• Created a program to identify uncommitted work and regularly email each analyst to save their code on a remote server. This helped protect the department from accidental code/resource deletion and ensured better standardization/formatting of code with automated linting and testing. • Developed an application for replacing the paper terminal logs. The tool cleanly integrates existing operations and maintenance data source, and other users’ data to help supervisors monitor and quickly restore rail service when issues occur and improve supervisors’ communications from terminal to terminal which has eliminated long-term issues with the rail service reliability, providing a better service to the riding public.• Created and automated multiple reports for Maintenance Departments that are used to perform Preventative Maintenance of CTA vehicles. This automation saves ~2 hours per day of managers’ time that can now be used to perform their core duties. In addition, the project has also helped address concerns about data validity and gave me a deeper understanding of the maintenance database, allowing me to help others in the Performance Management department to make better and deeper use of the data.

• Created a framework for quickly uploading data to our Redshift database, increasing data report reliability and reducing workload across the department. The framework was coupled with a site for an executive overview so individuals could monitor, at a glance, the status of all Performance Management uploads.• Created numerous automated reports designed to both supplement PM meetings and to give Garage Managers easy access to day-to-day information. In the field, these reports have improved vehicle cleaning across the fleet, allowed managers to discipline more drivers for short notice absenteeism, and reduced the number of buses being held for maintenance on the streets, improving service.• Introduced the use of larger datasets and more advanced statistics to the department. This includes a revenue report that uses multi-level seasonal adjustments to demonstrate trends in revenue and ridership, as well as automatically highlighting outlier routes and stations. In addition, a report was created that allows a manager to view an employee’s entire history, including weekends, vacation days, FMLA, and unexpected events such as Road Calls and discipline. This report enables managers address previously unnoticed trends in employee performance and absences.• Created a model using neural networks which uses garage and individual characteristics to predict how many and at what time absences will occur in the next day. This allows managers to decrease the number of lost runs without increasing the cost of manpower.• Shared my programming knowledge to support other members of the department. I used U+200B to help others with formatting in PowerBI, scraped websites for the daily Flash report using Selenium, and recovered lost credentials by changing password fields to text.

• Created product software for automated reporting and predictive failure models incorporating machine learning methodology.• Automated data gathering to monitor train schedule accuracy and passenger load reporting.• Using Twitter and Facebook API's, generated automated customer satisfaction reports through social media engagement.

Undergraduate position in research group led by Steven Levitt (author of Freakonomics).• Collaborated with graduate students, responsible for project delivery under tight timelines.• Using the programming language Python, created tools that organized and documented experimental data used in group research.• Optimized data acquisition to improve the speed of data entry and reduce database errors.

Jeff Metcalf Internship in team responsible for:• Designing and distributing promotional literature for large college events involving >100 alumni/faculty and >1000 students.• Contacting and persuading alumni and faculty to participate.• Mitigating last minute issues at events.• Reconciling budgets with planned activities.• Addressing questions and concerns of parents.

• Collated and analyzed academic articles for use in literature reviews.• Independently automated data entry from 2 minutes per manual entry to ~6 seconds with minimal supervision. • Created programs to collate and analyze data using R and Stata programming languages.• Helped organize and worked on a number of field experiments between September 2015 - November 2015 which involved door to door consumer surveys.• Participated in weekly research meetings with faculty and graduate students.

Primary contact for guests arriving at the Roosevelt Lodge in the evening and overnight.• Addressed questions and concerns of visitors.• Watched out for wild animals and called park rangers as needed.• Cleaned and maintained lodge public facilities.• Created Excel spreadsheet to explain employee time and activities to support facility budget.• Willingly accepted the jobs nobody else wanted to do (cleaning bathroom grout, etc.).• Worked up to 6 days/60 hours per week.