Riski Eka@forumiggcom95251

Flash sales and limited-stock deals are scattered across dozens of e-commerce pages, and tracking them manually means missed savings and stale information. DealScout AI is an agentic web-scraping assistant that hunts down requested deal information automatically: given a target product or category, it browses listings, extracts structured data (price, discount, stock status, seller rating), ranks the best offers, and pushes a concise alert. Under the hood, DealScout combines a tool-calling agent loop with an open-source LLM served on AMD Developer Cloud using ROCm/vLLM, demonstrating prompt engineering and multi-model serving on AMD GPUs. Planned extensions map directly to the AI Academy pathway: OCR reading of screenshot-based promos, RAG over product specs to ground recommendations and reduce hallucination, and a multi-agent setup where parallel scraper agents report to a coordinator agent. The project is built to be genuinely useful — it doubles as a real price-monitoring assistant for tech enthusiasts — while showcasing practical agent engineering, latency/accuracy trade-offs, and cost-aware serving on AMD hardware.

Six Matches. One leaderboard. Build on AMD, earn points and compete for prizes.