CivicLens

Inspiration Public spending and government datasets are often released in formats that are hard for ordinary citizens to understand. CivicLens was inspired by the need for data transparency and public trust — to help anyone, not just data scientists, explore how public money is used. We wanted to make open data truly open, using AI to bridge the gap between complex spreadsheets and meaningful insights. What it does CivicLens is an AI-powered transparency assistant that transforms raw government data into easy-to-understand insights. Upload a dataset or connect to an open data API. Gemini AI analyzes spending trends, flags anomalies, and answers natural language questions. Users can visualize trends, compare departments, and download reports directly from the dashboard. Designed for citizens, journalists, and policymakers alike — all powered by serverless infrastructure on Google Cloud Run. How we built it Frontend: Built with Next.js, offering a clean dashboard for uploads, queries, and visualizations. Backend: FastAPI running on Cloud Run, handling API requests and AI summarization tasks. AI Engine: Gemini 1.5 Pro in AI Studio generates insights, summaries, and explanations of public datasets. Data Pipeline: Uses Pub/Sub for async task orchestration, Firestore for session state, and BigQuery for structured data analysis. CI/CD: Automated deployment via GitHub Actions and Terraform for reproducible Cloud infrastructure. Challenges we ran into Parsing diverse data formats from different governments was tougher than expected. Balancing prompt determinism and creativity in Gemini outputs required fine-tuning seed values. Integrating multiple Google Cloud services while staying fully serverless pushed us to optimize API latency and container cold starts. Accomplishments that we're proud of End-to-end serverless architecture that scales automatically. Real-time AI explanations and anomaly detection powered by Gemini. Accessible UI that empowers non-technical users to explore civic data meaningfully. Fully automated build → deploy → inference workflow using Cloud Run and AI Studio. What we learned How to design prompt templates for reproducible AI outputs within data analysis tasks. Efficiently combining Gemini, Cloud Run, and BigQuery for near-real-time data summarization. The importance of UX clarity when presenting complex financial data to everyday users. What’s next for CivicLens Add multilingual support using Gemini’s translation APIs to reach non-English audiences. Extend to mobile and SMS-based civic queries for accessibility in low-connectivity regions. Integrate real-time policy dashboards for governments to communicate budget changes transparently. Explore AI-powered verification agents to detect and explain inconsistencies in public records.

Built With

• all