Helping communities mitigate hazards and plan for climate resilience

Developing geospatial analytics capabilities for Foursquare Studio, allowing customers to pull in data in all formats into a cutting-edge visualization and exploration tool. Contributing to the entire software stack from backend data processing to frontend interfaces to compute infrastructure configuration.

Worked on the Raster team, designing and building a complete raster data stack for elevation data, satellite/aerial imagery, and weather data. Built and maintained high performance processing code to ingest data and low latency highly-available APIs to serve it.

Integrated many different sources of data into pipelines to build and evaluate sophisticated data simulations for city transportation networks. Built tooling to verify data integrity and coordinate large and complex computations at scale.

Projects

Data-Product ETL Pipeline:

Designed/implemented an ETL pipeline to move generated models from raw form in Google BigQuery to a fast production database backing the end-user platform. Used Prefect to manage the dataflow DAG which ran on Dask/Helm/K8S.

Artigraph:

Helped design and begin implementation of a data dependency and metadata tracking library. Rather than being computation-centric like most dataflow libraries (Airflow, Prefect), this is intended to be data-centric. It focuses on the semantic relationship between different artifacts, abstracting out (and providing implementations for) different storage backends, file formats, and in-memory representations.

Helped develop and grow new machine learning solutions for cybersecurity and provide input on how to better manage and organize data for ease in analysis to improve overall efficacy

Projects

Automated Phishing Scam Detection System:

Helped deploy and evaluate Random Forest based detection system to identify credential phishing at very high accuracy rates and very low FP rates

Threat Timeline Analysis:

Developed infrastructure and data storage schema to correlate many internal events to better gain insight into performance and efficacy of legacy systems without adequate prebuilt monitoring.