Inspiration
Business decision-makers often struggle to extract timely insights from raw relational databases without relying heavily on technical data teams. DataPulse bridges this gap by translating plain English questions into executable, verified SQL pipelines paired with strategic executive insights.
What it does
DataPulse connects directly to structured databases, validates user intent against existing schema, executes deterministic SQL queries, and generates strategic narrative summaries along with interactive visualizations.
How we built it
- Google Gemini API (gemini-3.6-flash): For natural language query orchestration, schema mapping, and synthesizing raw data into executive insights.
- Streamlit & Plotly: For the interactive frontend UI and dynamic chart generation.
- Python & SQLite: For database connection, schema validation, and pipeline execution.
Challenges we ran into
Handling deterministic SQL translation while ensuring the model adhered strictly to table schemas without hallucinating column names.
Accomplishments that we're proud of
Building a working end-to-end multi-agent workflow that takes natural language, queries a live database, and delivers structured business insights within seconds.
What we learned
Optimizing system prompts for Google Gemini models to extract clean, executable code and professional-grade analytical summaries.
What's next for DataPulse
Expanding database support to multi-cloud enterprise data warehouses (Snowflake, Google BigQuery) and adding voice-activated analytical query capabilities.
Log in or sign up for Devpost to join the conversation.