Demo build · Custom LLMs & Automation

AI Product Research Analyzer for Resellers

A web app that researches a product across six marketplaces and search-trend sources in one run, scores demand, margin and competition, and gives a plain-language buy or skip verdict written by a language model.

The problem

Resellers deciding what to stock check prices and demand by hand across several marketplaces, then guess at margins. Supplier offers usually arrive as messy chat posts, not structured data, so even comparing options takes hours.

How it works

Structured data from messy posts. A language model turns unstructured wholesale posts into rows: product name, price, MRP, minimum order quantity and category.
Live multi-platform research. Browser automation checks six marketplace and trend sources, captures screenshots as evidence and falls back to a manual-check link when a site blocks automation.
Scoring. A transparent scoring model rates demand, margin, customer satisfaction and competition, and estimates retail median, wholesale cost and profit.
Verdict and report. A language model summarises the findings into a short verdict, delivered as an HTML report with the screenshots attached.
Multi-user app. Login, job history and a queue that runs one research job at a time.

What it demonstrates

LLM extraction for turning unstructured text into reliable fields, combined with deterministic scoring the user can inspect, so the AI explains the numbers rather than inventing them.