About me
Using data driven insights, I help business leaders answer difficult questions by translating big data into actionable information. I have done work for clients across many industries like finance, politics, manufacturing and government. I deliver value to my clients by driving innovation and pushing creative solutions to problems.
I originally studied Finance and Accounting and then moved into the Data Science field. I have been able to combine both experiences into a full stack of sophisticated data and financial analytics. Since then, I have been working as a Management Consultant where I provide clients with business insights to help them achieve their financial and internal objectives.
Working with talented individuals and accomplishing their goals is something I’m truly passionate about. Being able to creatively problem solve and face new challenges every day is what makes Data Science so rewarding.
Curious how you can capture value from your data? Let’s chat.
- Email : thomasbacas@gmail.com
- Location : Arlington, VA.
- Resume : Download
- Phone : +1-571-230-4773
My services
PYTHON DEVELOPMENT
Customized tools and automation to help your teams work faster and smarter.
PREDICTIVE ANALYTICS
Utilize machine learning models to translate big data into actionable information.
DATA ENGINEERING
Put your data at your fingertips with efficient pipelines and data stores.
BUSINESS INTELLIGENCE
Understand insights from your business processes with ad-hoc data analysis and easy access dashboarding.
CLOUD COMPUTING
Capture the cost savings and scalability of moving to the cloud with services like AWS, Azure and GCS.
TECHNICAL STRATEGY
Not sure where to start? Let's talk about how to move your business forward by capturing and utilizing your data.
My Portfolio
NEURAL NETWORKS AND PREDICTIVE MODELING
Neural Networks, Random Forest, Ordinary Least Square Regression, Python
Unlike the stock market, the bond market has uncharacteristically less information available for traders. Equity traders have current and historic trade data available quickly and cost effectively while Bond traders suffer from information drought. This leads to bond prices being significantly less efficient than those of equity market. With this in mind, this project sought to develop a model to produce more accurate pricing predictions for bonds. Utilizing data from Benchmark Solutions and Neural Networks/Ensemble models, this product can predict accurate and computationally effecient bond prices.
ZILLOW HOUSING VALUATION FOR DAMAGE ESTIMATES
Python, Geopandas, Zillow Api, MySQL
New Light Technologies, a contractor for FEMA, requested dynamic code for producing housing estimates for a given zipcode. During a disaster, it is important to model and estimate the potential or forecasted effect of the event, including the projected/forecasted damage. FEMA has already developed estimates for other factors. This project sought to add an additional indicator, up-to-date estimates for property value in the area of effect. Utilizing Zillow API, this product allows for quick return on property statistics of an impacted area or zip-code.
REPORT AUTOMATION
Python, AWS, Django, MySql, Openpyxl
After generating predictions from sophisticated machine learning models, the client wanted simple to read and easy to digest Excel reports. To manage this, Python tools were created to automate the report generation process, as well as store and encrypt data on the AWS Cloud. This project gave the client the tools to create these reports automatically with a dynamic web interface designed in Django. The result was substantial time savings, from reports taking 3 days to finalize initially to 2 hours after implementing the app.
PREDICTING VOTER SEGMENTATION
Python, AWS, Django, MySql
In order to maximize their strategic positioning, the client sought advanced analytics with insights into the specific trends of population responses to niche issues and political races. To create the most efficient and cost-effective solution for the client, multiple instances of classification and regression models were executed at scale. Over 50 models (trained on real survey responses) were run asynchronously via AWS EMR, which quickly processed hundreds of gigabytes of data to generate predictions of voter turnout and segmentation. These predictions were utilized by the client to influence their targeted marketing efforts and inform individual political campaigns on shifting voter developments.
FEC DATA ANALYSIS AND ENGINEERING
Python, AWS, PostgresSql, Classification and Regression Modeling, R
With hundreds of gigabytes of publicly available funding data brought in by the FEC, the question is asked “how can we use this to better understand donors”. This project sought to translate billions of FEC transactions into data points that can be used to better understand donor behavior. This problem was multistage. The first step was cleaning messy FEC data effectively, next was migrating that data to a cloud data warehouse where it could be added to a suite of models already processing other consumer data. With AWS tools, an automated pipeline was developed in Python to pull down files daily from the FEC website, this pipeline would perform standard cleaning ETL steps and move them to an AWS MySql database where it would be joined to consumer data and pushed into model predictions. Not only did this process aid the client in improving model predictions but as a result of readily available FEC data, they were able to produce weekly newsletters on FEC transactions preceding the 2020 presidential election.
FINANCIAL HEALTH AND SYSTEM MIGRATION DASHBOARD
OracleDb, Python, PowerBI, Openpyxl
The Postal Service processes nearly 200 million pieces of mail a day. Validating the postage, weight and dimension of each item of mail is an essential part of their business. The various finance stakeholders at the Postal Service needed an automated, easy to digest dashboard to understand the effects of daily changes of postage refunds, adjustments and system migrations. Many pieces of data came from multiple stages of the enterprise system, while not siloed, these datapoints had no channels of easy movement. It was clear that reinventing the wheel would not be an option for connecting these disparate teams. An intricate set of Python scripts was developed to collect the various datapoints in one central database, which connected to a PowerBI dashboard and presented the data to the financial stakeholders.A similar dashboard was requested by the USPS to fully monitor their system migration. This project would be used to distribute information all throughout the Postal Service’s MEPT leadership, and help inform multiple teams of financial shifts, daily errors, and necessary system adjustments.
Testimonials
Thomas worked on my team of data science consultants for about a year and he consistently produced excellent work. Regardless of whether the project was analysis or engineering-based, Thomas used his coding skills and careful judgment to meet our clients' needs. If you are considering hiring Thomas as a freelance contractor, I would highly recommend his services.
Jake Russ
Chief Data Scientist