Inspiration The inspiration for 311 Insights Engine came from the sheer volume of NYC Open Data. With millions of 311 requests filed annually, the "signal" often gets lost in the "noise." We wanted to move beyond simple bar charts and create a tool that actually thinks like an urban planner—identifying not just what is happening, but providing a roadmap for how to fix it. What it does 311 Insights Engine is a real-time diagnostic dashboard for New York City. Users can filter by borough to see a live snapshot of the top 1,000 most recent service requests. The app automatically identifies the most prominent issue (e.g., Illegal Parking, Heat/Hot Water) and uses the Gemini 1.5 Flash API to interpret that data, offering strategic policy suggestions and resource allocation advice.
How we built it Backend: Python and Pandas for high-speed data manipulation and filtering of the NYC Open Data SODA API.
Frontend: Streamlit for a clean, responsive, and interactive web interface.
Intelligence: Google Gemini 1.5 Flash API to perform semantic analysis and generate actionable urban insights.
Data Source: Live 311 Service Requests from the NYC Open Data portal.
Challenges we ran into The primary challenge was the short time constraint. We had to balance the massive scale of the NYC 311 dataset with the need for low-latency AI responses. To solve this, we implemented smart data "slicing" to ensure the AI received a high-quality, relevant snapshot of the most recent data without hitting rate limits or slowing down the user experience.
Accomplishments that we're proud of We are proud of successfully bridging the gap between raw data and human-readable strategy. In just 180 minutes, we went from a blank script to a working prototype that doesn't just display numbers, but actually suggests how to improve the quality of life for New Yorkers.
What we learned We learned how to leverage Large Language Models (LLMs) for specialized data analysis. We discovered that by providing the AI with specific data summaries (like value counts of complaints), it can offer surprisingly nuanced suggestions—such as recommending specific types of infrastructure inspections based on a cluster of seemingly unrelated complaints.
What's next for 311 Insights Engine Predictive Analytics: Using historical data to predict which neighborhoods will see spikes in specific complaints (e.g., predicting heating issues before a cold wave).
Multimodal Analysis: Integrating Gemini’s ability to "see" by analyzing photos attached to 311 tickets to prioritize urgent repairs (like dangerous sinkholes).
Citizen Feedback Loop: A feature allowing residents to see how the AI is advocating for their specific neighborhood issues.
⚙️ Built With Python (Core Logic)
Streamlit (Web Framework)
Pandas (Data Processing)
Google Gemini 1.5 Flash (AI Engine)
NYC Open Data API (Data Source)
Log in or sign up for Devpost to join the conversation.