We are building An AI-powered tool that lives on a website where visitors enter their company name, the system matches against a proprietary database, pulls their data and peer benchmarks from APIs, feeds everything into Claude API with a purpose-built system prompt, and renders a custom narrative analysis instantly on the page. Every completion captures a lead and syncs to HubSpot. You are also building an automated report engine that generates branded PDF outputs from the same data sources. Your stack: - Claude API (Anthropic), n8n, React or HTML/JS, HTML-to-PDF rendering, HubSpot Scope (3 components): - AI audit tool (the main build): Lightweight frontend (React or clean HTML/JS) with company name autocomplete field and email capture. n8n backend receiving the form submission, querying two external data APIs for company-specific information and peer benchmarks, assembling the data payload, calling Claude API with a system prompt and few-shot examples (we provide the prompt), parsing the structured JSON response, and returning it to the frontend for rendering. Below the results: a booking CTA linking to a calendar page. - Every completion syncs lead + audit data to HubSpot via n8n. Must handle edge cases: no company match, incomplete data, repeat visitors. - Report engine: n8n workflow that queries the same data APIs, passes to Claude API for narrative generation, and renders a branded 1-2 page PDF via HTML-to-PDF. Triggered by form submission or engagement signal. Template-based: we provide the brand assets and layout spec, you build the rendering pipeline. - Inbound conversion flow: post-audit email nurture trigger in HubSpot (if they don't book). Auto-delivery of PDF report for booked meetings. Pre-call briefing card auto-generated for the sales team via n8n and HubSpot. Required (non-negotiable): - Production experience with Claude API or similar LLM APIs (OpenAI, etc.). Not just chatting. You must have built applications that call the API programmatically, handle structured JSON output, and manage prompt + few-shot context. - Frontend development (React or clean HTML/JS/CSS). You need to build a fast, clean, branded page that renders dynamic content from an API response. - n8n experience (multi-step workflows with API calls, conditional logic, error handling). - API integration (REST APIs, webhooks, JSON transformation, error handling). To apply, answer these 3 questions: - Describe a production application you built using Claude API or a similar LLM API. What was the use case, how did you handle the prompt/context, and what was the output format? - Show or describe a web application you built where a user inputs data, it calls external APIs in the background, and renders dynamic results on the page. How did you handle loading states, errors, and edge cases? - Have you built any automated PDF generation pipelines? What tools did you use and what triggered the generation?