ShedTrace

About the Project

ShedTrace was inspired by a simple question: Why is a sidewalk shed still standing, and what does the public record actually say about the building behind it? In New York City, the information needed to investigate a building already exists, but it is scattered across multiple public datasets. We wanted to turn that disconnected information into something people could actually understand. This led us to build ShedTrace, an evidence tool that connects sidewalk shed history, DOB violations, DOB complaints, elevator safety records, and FDNY fire inspections through the building’s BIN (Building Identification Number).

We built the project using Python and pandas to download, clean, process, and join the datasets, Streamlit for the interactive interface, Google Gemini for a building specific AI assistant, and Docker/Docker Compose for our development environment. When a user enters an address, ShedTrace identifies the building, finds its BIN, confirms an active shed, and connects records from the five datasets into a chronological building profile. We also created a 3D visualization to make the volume of violations and complaints easier to explore.

One of the biggest things we learned was how much work goes into turning public data into useful information. Finding the datasets was only the beginning; we had to understand different schemas, clean inconsistent records, connect everything using BINs, and translate technical codes into language that is easier to understand. We also learned how to combine a data pipeline, interactive web interface, visualization, and AI feature into one application under a very limited hackathon timeline.

Our biggest challenges were working with multiple datasets that were not designed as one system, keeping the data relationships accurate, and making sure the AI assistant stayed grounded in the building’s actual records rather than guessing. We also had to make intentional scope decisions. Instead of building a generic NYC building search tool, we focused specifically on buildings with active sidewalk sheds because our core question was: why is this specific shed still standing?

ShedTrace is currently based on a development snapshot of NYC public data, but the project could eventually support live data updates, citywide building lookup, historical elevator tracking, and a multi-building 3D map. Most importantly, this project showed us that meaningful tools can be built by taking information that already exists and connecting it in a way that makes the evidence easier to explore and understand.