Shoppers who want the best price for a product must visit several e-commerce websites, search for the same item on each, and compare prices, offers and ratings by hand. This is slow, error-prone, and complicated by the fact that different retailers describe the same product with different titles. This paper presents AI PriceWise, a full-stack system that accepts a product query, collects live listings from Amazon, Flipkart and Reliance Digital using Playwright browser automation, normalizes the scraped data, and groups listings of the same product using a hybrid matcher that combines Sentence-Transformer semantic similarity (all-MiniLM-L6-v2) with brand, model and specification scores and hard variant, storage and quantity gates. For each matched product the system computes lowest, highest and average prices, savings and unit prices, classifies the current price as a good deal, fair price or high price against the product's own observed price history, predicts a near-term price range with an XGBoost regressor when enough history exists, and produces a grounded natural-language explanation with a Groq-hosted large language model. The backend is built with FastAPI and SQLAlchemy and the frontend with Next.js. The system was verified with an automated suite of 213 passing tests, 30 documented functional test cases and manual live searches; in a live search for iPhone 16, listings from all three stores were matched into one product, and when one store timed out the system still returned results from the remaining stores.
No takes yet. Share an insight, caveat, or question.
Priyadarshini et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: