Projects

Flamcon: Flamingo+Falcon Language-Video Cross Attention Network
Flamcon is a flexible and powerful deep learning architecture that enables efficient cross-modal attention between language, image, and video data. The repository contains the code for training and using Flamcon, along with detailed descriptions of its components and how to get started.
‍
(pytorch)
git
MunicipalAI
MunicipalAI is an open source project dedicated to leveraging the power of automation to minimize waste in government operations. We believe that by streamlining processes and employing advanced machine learning techniques, we can significantly reduce inefficiencies and save taxpayer funds.
‍
(react native, ios, android, django, pytorch, transformers, c++, jetson nano)
git
Fire Overtime Optimization
Used reinforcement learning to prove that an algorithm could optimize overtime.
‍
(XGBoost, MuZero, python, c)
git
Tulsa Equality Indicators
Partnered with the CSC and SUNY to collect and validate data for the Tulsa Equality Indicators reports (2019,2020). Wrote data governance and etl process to automate extraction of city data each year.
‍
(SQL Server, SQL, Dremio, Sharepoint Workflows)
Eviction Prediction
Using lstm's and deep learning I used time-series data to predict eviction. Project was used as an example of Tulsa's innovative work for the Prime Minister of UK.
‍
Special thanks to TU Student Jiana Meng and BIT for peer reviewing, duplicating research, and improving the address matching algorithm.

Also, thanks to Asemio and OK Policy Institute for providing eviction data.
‍
Finally, thanks to Hannah Ralston for helping run the RCT with BIT.

(python, keras, lstm, sql, text messaging)
What Works Cities
Data Governance
Established a team of twenty data stewards who helped implemented Data Governance and improve Open Data across fifteen departments at the City of Tulsa. Leading to improving the WWC Data Governance and Open Data Certification score by a cumulative 80%, twelve departments using B.I. and hundreds of thousands of dollars in savings(per water and sewer).
‍
(sql server, dremio, tableau, python, r)
BloombergNews Channel 8 Tulsa
Mayor's Dashboard
Developed a KPI dashboard based on the top three metrics for each of the fifteen departments. Data refreshed weekly.
‍
(sql server, dremio, tableau)
Solar Powered AI Camera
Used custom built solar powered camera and ai to count usage of tennis courts.
‍
(api, python, dnn, opencv, raspberry pi)
Aerial Instance Segmentation
Converted 200 square miles of tiff images to coco dataset. Trained ai model to detect impervious services.
‍
(python, coco, masked-rcnn, opencv)
The hardest data science tournament in the world.
Using numerous advanced ml techniques I finished in the top 5%.
(python, ml, stats, data science)
Urban Data Pioneers
Led cohorts 3-7 of the award winning program.
City of Tulsa UDPHarvard Data-Smart City SolutionsCities of Service Blueprint
Tulsa Data Science
Founded a non profit with the goal of using data science to benefit Tulsa.
Our partnership with the City of Tulsa led to being mentioned in a TED talk and helping win the engaged cities award.
Engaged Cities AwardTED Talk
CPC+ Engine
Built a CPC+ measure engine for the Ohio Health Collaborative. The engine processed 250 million records, meeting evolving company standards.
(sql server, sql, nifi, java, python)
Fortune 1000 Data Governance Project
Member of a three month data strategy project.  Performed qualitative assessment -through interviews and surveys- of how the company used and wanted to use data.  Developed a long-term strategy for the company.  Strategy was reviewed by a third party and we were told it was well done.
‍
(Qualitative and Sentiment analysis, Surveys, Interviews)
Fortune 1000 Business Process Improvement
Led project to convert paper forms -in the field- to digital. Allowing the company to implement consistent KPI’s and management control systems. The architected solution was a responsive web and IOS and Android mobile application. Delivered a stable production product in less than two months. Built to work in remote environments with intermittent internet connections. Used mostly Open Source software. Cross trained developers to support the application. Was in production for over three years. Project was estimated to save the company $3,000,000. (Per Renoir Consulting)
‍
(Ionic, nodeJS, AngularJS, JSON, HTML5, CSS, SQL SERVER, OKTA, REST)
Fortune 1000 Root Cause Analysis
Member of a three month long engineering project. Performed feature engineering to look for contributing factors to a customer complaint. Built stored procedure to allow smoothing of dependent variables in historic data. Also built a dashboard and real-time tool to allow monitoring the issue.  Used univariate analysis, ridge and lasso regression, and random forests to look for correlations. Wrote a paper to document my findings. Project was deemed a success.
‍
(Microsoft APS Stored Procedures, Tableau, Python, Sklearn, Euler-Legrange Equations of Motion)
Fortune 1000 Data Engineering Project
Wrote a multithreaded C# MSMQ windows service in two days that replaced an over-engineered three month long project. Service was used to process all the companies field data. It was was used in production for three years with 99.99 percent of the data processed successfully.
‍
Rewrote the service as a JAVA nifi processor which was also used in production.
‍
(C#, JAVA, NIFI, Windows Service, Data Engineering, MSMQ)
Fortune 1000 Real-time Prediction
Engineered FLINK tool to perform real-time engine overheating prediction. Collected real-time weather information from an API and used proprietary real-time data to create a combined real-time dataset. Applied a linear regression model to the dataset to predict engine overheating and to warn the company. The tool minimized downtime, ultimately saving the company money. Updated version was in production for over a year.
‍
(FLINK, JAVA, Kafka)