DataScout
Agentic AI Data Analyst
Ask a dataset a question in plain English. An agent writes the Python, runs it in a sandbox and shows its work.
- 01User
- 02Natural language
- 03AI agent
- 04Python execution
- 05Analysis
- 06Insight
Which region grew fastest quarter-on-quarter?
df = load("sales.csv")
q = df.groupby(["region", "quarter"])["revenue"].sum()
growth = q.groupby(level=0).pct_change()
growth.groupby(level=0).last().idxmax()
→ South leads — computed, not guessed. Code attached.
- Context
- AI for Bharat · Feb – Mar 2026
- Role
- AI Engineer (Developer) · team lead
- Cloud
- AWS serverless
- Team
- 4
01 Problem
Most people with a business question can’t write the pandas to answer it — and a chatbot that predicts numbers is not the same as one that computes them.
02 Approach
Ask in plain English
Upload a dataset and ask a question in natural language through a Streamlit interface — no SQL required.
An agent plans the analysis
An agent on Amazon Bedrock interprets the analytical intent behind the question.
Generate and execute Python
The agent writes Python that runs in a sandboxed environment, so results are computed rather than guessed — and the code is there to audit.
Serverless on AWS
S3 for datasets, Lambda and API Gateway for execution, DynamoDB for state.
Secure by design
IAM role isolation, AES-256 encryption, audit logging and sandboxed execution.
03 Results
- Team — led
- 4
- Encryption at rest
- AES-256
- SQL required
- 0
- Reached the Prototype Development Phase of the AI for Bharat national AI hackathon.
- Led a team of 4 through the build.
04 Stack
- Amazon Bedrock
- AWS Lambda
- S3
- API Gateway
- DynamoDB
- Streamlit