Recidiviz is building an open-source, technical infrastructure to help safely, sustainably, and equitably shrink the criminal justice system.

Lead technical efforts at Omelas—designing and managing implementation of high-scale real-time machine learning systems. Projects included topic modeling of propaganda, automatic quote extraction and attribution of world leaders, entity relation of geopolitical entities, neural translation of text, and others. I also mentored members of the data science and engineering teams, and wrote technical language of grant proposals for government clients.

Built and maintained a cloud-native data ingestion system with python, Docker, Postgres, Kubernetes, and AWS. Lead an engineering team ingesting hundreds of thousands of pieces of propaganda a day across social media and online news sources.

Lead graduate fellows and in-house technical staff at US-SOCOM in the development of a machine learning system to triage media analysis for counter-terrorism efforts. Model development used python, Tensorflow, and OpenCV. I briefed four star leadership for the development and handoff of the project.

Evaluated the impacts of land use and reform policies in post-WWII Italy through a spatial regression discontinuity design using R, python, QGIS, and ArcGIS.

Conducted quantitative research at the UChicago Consortium on School Research focusing on the implementation of Common Core State Standards within Chicago Public Schools. This research identified the relationship between teachers’ reported instructional practices, the district’s capacities to enhance instruction, and students’ classroom experiences and performance using hierarchical linear modeling and a decade of surveys and administrative data. Beyond research, I provided training to Consortium staff on bash, git, R, SQL, and general best practices for technology use in research. Analyzed the implementation of Common Core State Standards in Chicago Public Schools with Hierarchical Linear Modeling and other Econometric methods in HLM5, SAS, and R.

- Developed and refactored the iOS applications FaceOff and genmo, mobile applications deploying live machine learning models, in Swift and xcode.- Lead and managed development projects resulting in a modular codebase to be used across applications in git and GitHub.- Redesigned and simplified the company website: https://latentculture.com

Instructed Thinkful students on: - computational thinking and programming in python- the python data science stack (pandas, numpy, matplotlib, sklearn)- SQL (postgres)- statistics and inference (descriptive stats, hypothesis testing, regression)- supervised learning - unsupervised learning- data science problem formulation and evaluation

Created a website and resource for sharing policy-relevant datasets and code that have been accumulated and used by researchers within the University Community. I handled both the structure and front-end of this website, interviewed researchers on their work with the data, and constructed use cases for each dataset in R and python. Tools: Jekyll, Bootstrap, R, python, scss

Assisted instruction of graduate students learning data collection, manipulation, and analysis through Python, SQL, git, and bash.

Developed and implemented the Auto Bodily Injury Claim SMART Model for Complexity Adjusted Benchmarking - extending the award winning SMART paradigm. Specifically: - Provided insights on claim complexity through causal inference / econometric analysis in R and python- Predicted complexity targets with various supervised learning methods (Sci-Kit Learn)- Classified claim complexity groups with multiple supervised learning methods (Sci-Kit Learn)- A high level geospatial analysis of business (QGIS, SQL Server spatial)- Deployed the trained model and integrated with automated reporting (python, SQL Server, Excel)

Constructed a ‘Plant Based Database’ in Excel to provide data driven insights into the sustainable food market. Product data was obtained by communicating directly with companies and scraping public websites using BeautifulSoup and urllib3.

Reduced utility payment time by 80% and improved accuracy through automating processes with VBA. Managed over 250 utility accounts by integrating Excel with the TAMU-C Financial Accounting Management Information System.

Composed, arranged, and rehearsed music to facilitate a learning experience for young people.

Assisted with the study of White Dwarf stars by constructing a data pipeline. Given star coordinates and CCD images, the pipeline obtained data with IRAF (Image Reduction and Analysis Facility) and stored relevant information as a csv in a Fedora Linux environment.