AI Agent Data Analysis – Mobile Sales Dataset

Project Overview

This project demonstrates how to build a lightweight AI-assisted data analysis pipeline using Python and the OpenAI API.

Instead of sending the full dataset to the AI model, the system generates a structured metadata summary (schema, statistics, missing values, etc.) and sends only that summary for analysis. This approach:

• Reduces token usage
• Improves performance
• Minimizes unnecessary data exposure
• Maintains scalability

The project simulates an AI-powered data analyst that reviews dataset structure and produces business insights.

---

Dataset

Source: Kaggle
Dataset: Synthetic Mobile Sales 2025

The dataset contains simulated mobile device sales data, including:

• Product information
• Sales quantities
• Revenue
• Transaction dates
• Regions
• Additional transactional attributes

---

Project Architecture

The workflow follows this structure:

1. Install dependencies
2. Download dataset from Kaggle
3. Load dataset into Pandas
4. Generate automated metadata summary
5. Send structured summary to OpenAI
6. Receive AI-generated business insights

Only summarized metadata is transmitted to the model.

---

Step 1: Environment Setup

Install required dependencies:

pip install kagglehub[pandas-datasets] openai pandas numpy

Import required libraries:

import os
import json
import pandas as pd
import numpy as np
import kagglehub
from getpass import getpass
from openai import OpenAI

Securely load your OpenAI API key:

os.environ["OPENAI_API_KEY"] = getpass("Enter your OpenAI API key: ")
client = OpenAI()

Important: Never hard-code API keys into a public repository.

---

Step 2: Download Dataset from Kaggle

Programmatically download the dataset:

path = kagglehub.dataset_download("syedaeman2212/mobile-sales-data")

---

Step 3: Load Dataset

Load the CSV file into a Pandas DataFrame:

df = pd.read_csv(os.path.join(path, "synthetic_mobile_sales_2025.csv"))

Preview the dataset:

df.head()
df.info()

---

Step 4: Generate Structured Data Summary

Instead of sending raw records, build a structured metadata summary:

summary = {
    "shape": df.shape,
    "columns": df.columns.tolist(),
    "dtypes": df.dtypes.astype(str).to_dict(),
    "missing": df.isnull().sum().to_dict(),
    "numeric_summary": df.describe().to_dict()
}

Convert summary to JSON for API transmission:

summary_json = json.dumps(summary, indent=2)

This ensures:

• Reduced token usage
• Faster API responses
• Improved security
• Scalability

---

Step 5: Send Metadata to OpenAI

Construct a prompt that instructs the AI to analyze the dataset summary:

prompt = f"""
You are a senior data analyst. Based on the following dataset metadata, provide:

1. Key insights
2. Observed trends
3. Data quality concerns
4. Potential business recommendations

Dataset Summary:
{summary_json}
"""

Send request to the API:

response = client.responses.create(
    model="gpt-4.1-mini",
    input=prompt
)

analysis_output = response.output_text
print(analysis_output)

---

Step 6: AI-Generated Insights

The AI returns:

• Business insights
• Identified patterns
• Anomaly detection suggestions
• Data quality observations
• Strategic recommendations

This simulates an AI-powered analytical review.

---

Design Rationale

Sending only metadata rather than full datasets:

• Reduces API token costs
• Prevents unnecessary data sharing
• Improves performance
• Maintains governance standards
• Enables production scalability

This architecture is well-suited for enterprise AI analytics pipelines.

---

Potential Enhancements

The project can be expanded to include:

• Automated visualizations (Matplotlib, Seaborn, Plotly)
• Correlation matrix analysis
• Outlier detection
• Feature engineering
• Machine learning model training
• Automated report generation (PDF or HTML)
• Interactive dashboard (Streamlit)
• User-uploaded dataset support

---

Skills Demonstrated

• Data ingestion and preprocessing
• Automated exploratory data analysis
• Structured metadata engineering
• API integration with OpenAI
• Efficient token management
• AI-assisted analytics pipeline design
• Reproducible Python workflows

---

Production Considerations

For real-world deployment:

• Add environment variable management
• Implement error handling
• Add structured logging
• Implement API rate limit handling
• Containerize with Docker
• Deploy via a cloud platform (AWS, Azure, GCP)

---

Use Case

This project demonstrates how to build an AI-assisted analytics system capable of reviewing datasets and generating actionable insights without exposing raw data.

It is suitable for:

• AI-driven reporting systems
• Internal analytics assistants
• Enterprise data governance workflows
• Portfolio demonstration of applied AI integration