I want to build an AI-powered search engine that handles general web queries while surfacing the best user-generated content first. The core of the project is a full search stack—crawler, indexer, ranking model, and a lean front-end—that can ingest millions of pages, classify what is user-generated, and boost it in the results without sacrificing relevance or speed. Key goals • Crawl and keep an up-to-date index of the open web. • Detect forums, Q&A threads, community blogs and similar sources so they can be weighted more heavily than corporate or news sites. • Train or fine-tune an AI ranking model that understands conversation style, sentiment, and authority signals unique to user contributions. • Expose a simple web interface plus JSON API to test queries, inspect ranking explanations, and export results. Technical freedom is yours—ElasticSearch, Solr, Meilisearch, Vespa, or a custom vector store are all fine as long as they scale horizontally and let me tweak scoring features. Python, Java, Go or Rust for the backend; React, Vue or plain HTML/CSS for the UI—use what you are fastest with. Deliverables 1. Source code and deployment scripts (Docker or Kubernetes). 2. A live demo instance seeded with at least 1 M pages. 3. Documentation explaining architecture, schema design, ranking signals, and how to extend filters or UI components. 4. A short video walkthrough of the system in action. Acceptance will be based on: • Query latency under 500 ms on the demo set. • Clear promotion of user-generated links in the top 10 results for at least 80 % of provided test queries. • Clean, repeatable deployment from repo to cloud. If you have prior work on search relevance, NLP, or large-scale crawling, please share a concise example when you respond.