PR: Add Product Price Estimation Pipeline (Price is Right)

Overview

This PR introduces the Price is Right capstone project, a comprehensive machine learning pipeline designed to estimate product prices from text descriptions. The tool transitions from data exploration and baseline benchmarking to interactive performance visualization, utilizing the ed-donner/items_lite dataset to evaluate how LLM-based regression scales with training data.

Key Features

• Automated Data Ingestion: Integrated fetching and processing of the Hugging Face items_lite dataset, converting raw signals into structured pandas DataFrames for analysis.
• Baseline Benchmarking: Implements a statistical "Mean Baseline" model to establish a performance floor ($MAE$) against which all subsequent fine-tuned models are measured.
• Interactive Analytics Dashboard: A custom-built Gradio interface that allows users to toggle between price distributions, market segment breakdowns, and model scaling laws.
• Performance Scaling Analysis: Tracks and visualizes the "Learning Curve," demonstrating how Mean Absolute Error (MAE) improves as the number of training samples increases from 0 to 5,000.

What’s Inside

• Backend: Data pipeline using datasets and OpenAI SDK integration for future fine-tuning extensibility.
• Logic: Custom evaluation suite including a run_baseline_mean function and pd.cut categorization for market segmenting (Budget vs. Luxury).
• Visuals: A three-tier visualization suite using Seaborn and Matplotlib to render histograms, pie charts, and regression performance curves.
• Workflow: A streamlined Jupyter execution flow from .env configuration to a live-hosted Gradio dashboard.

Screens:

CC: @ranskills