Smart Research Dashboard – AI-Powered Decision Assistant

Our dashboard lets anyone upload an Excel or CSV file and instantly transforms it into insights. It automatically visualizes trends, answers natural language questions.

Comment

Inspiration

Making sense of raw survey or business data is time-consuming, and often requires specialized tools. Teams want actionable insights, not spreadsheets. We were inspired to build a tool that transforms everyday Excel files into clear decisions by combining AI summarization, visualization, and live benchmarks from the web.

What it does

Smart Research Dashboard is an AI-powered decision assistant.

📂 Upload an Excel/CSV file → the system ingests and indexes it in TiDB Serverless with vector search.

📊 Instantly generates visualizations like trend graphs and breakdowns.

🤖 Lets you ask natural language questions about the data and receive summarized answers via Hugging Face models.

🌍 Enriches results with free web search APIs, comparing your data to industry benchmarks.

📝 Produces clear summaries so users can focus on decisions, not raw numbers.

How we built it

Frontend: React (Next.js + Tailwind) for data upload, chart visualization, and chat-style Q&A.

Backend: FastAPI (Python) orchestrating data ingestion, querying, and AI calls.

Database: TiDB Serverless with vector search to store and retrieve embeddings of uploaded datasets.

AI Models: Hugging Face Inference API for Q&A and summarization.

External Data: Free web search APIs (Search1API, serpstack) for benchmarks and external insights.

Challenges we ran into

Efficiently converting uploaded Excel data into structured + vector embeddings.

Handling natural language questions that required combining internal data with external benchmarks.

Staying within the free tiers of Hugging Face and search APIs.

Designing a smooth frontend that balanced charting, Q&A, and summaries in one interface.

Accomplishments that we're proud of

Built a multi-step agentic workflow that chains TiDB, Hugging Face, and free web search APIs end-to-end.

Created a simple, intuitive interface where anyone can upload a spreadsheet and get insights instantly.

Successfully benchmarked local data against real-time external knowledge.

Proved how TiDB’s vector search can power RAG-style workflows with structured business data.

What we learned

How to design agentic pipelines that integrate multiple services (TiDB + LLMs + external APIs).

The importance of data preprocessing for embeddings when working with spreadsheets.

That even small free-tier APIs can be powerful if orchestrated cleverly.

How visualization and summarization together make AI insights easier to trust and act on.

What's next for Smart Research Dashboard – AI-Powered Decision Assistant

Expand visualization capabilities with interactive dashboards.

Add multi-file support (combine several datasets into one view).

Built With

Share this project:

Updates