Doing a little bit of everything in order to build a platform where you can find your next favorite thing.Web Application: managing and contributing to the technical effort of our frontend application, based on NextJS and TailwindCSS.Discovery Platform: develop the core backend API that surfaces relevant, interesting products from thousands of curated, smaller brands to end users. Manage distributed data processing via Argo Worfklows, BigQuery (and DBT). Additionally, train and deploy a variety of machine learning models, primarily written in PyTorch and making heavy use of Transformers architectures.

Manage the core engineering platform at Hawkfish, consisting of compute infrastructure, data engineering pipelines, and machine learning infrastructure.- Provision software infrastructure on Google Cloud Platform using Terraform for infrastructure-as-code- Build workflows around Data Build Tool (DBT) and BigQuery that democratize core aspects of the data engineering process- Provide reproducible and scalable machine learning infrastructure and a suite of diagnostic tools to data scientists

Managed a team focused on providing Machine Learning infrastructure to data scientists.- Delivered reproducibility to data scientists via a Kubeflow Pipelines cluster running on Google Kubernetes Engine and scalabality via Argo Workflows and Kubeflow Katib- Develop and maintain Python-based re-usable components that help data scientists quickly jumpstart new projects using scikit-learn, XGBoost, and PyTorch- Maintain a Model Quality Control Dashboard that provides a suite of model diagnostics to evaluate their models

Lead a team focused on bringing data science insights to political domain experts by building highly analytical web applications that enabled better scenario planning and decision making.

Half infrastructure developer, half data scientist. Started a new team focused on distributed data processing and anomaly detection.-Model time-series data using a variety of methods (ranging from MLE to GAN) to proactively identifyanomalies in the data.-Develop NLP models that automate the labeling of key clauses in Mortgage Backed SecurityProspectus documents.-Maintain and enhance a distributed data-processing system and migrate core components of this system to PySpark

-Developed Python scripts that automated a variety of tasks from PDF parsing to CSV extraction, transform, and loading. Improved the straight-thru processing percentage of a particularly nuanced subset of commercial loans by 400%, saving analysts over 15 hours of manual work per month.-Maintained and enhanced a Java Spring-Framework shell utility that is used by the analyst team to make data updates. Migrated legacy code to this new system, and created an interface for calling Python scripts from the Java codebase.-Built a Python tool that simulates user-input to Bloomberg CMBS analytics screens, allowing front-end developers to test the alpha and beta environments versus production for stability and data integrity.

-Used R and python as a member of a four-person Think Tank to identify novel content on Twitter, and presented this concept to the head of Global Data as a business proposal. Used the R software suite to conduct statistical analysis on Twitter metadata and developed an algorithm that accurately triggered relevant Twitter alerts.-Configured automated parsing of time series data-Published fair value prices to {BOIL} in order to bring transparency to the physical refined products market.