Build a Dockerised LLM Website Evaluator Service

We are looking for a strong full-stack / backend contractor to build a standalone HTTP service that evaluates a prospect’s website and determines what technical capabilities we would need to service it. This is not a generic scraper. It needs to behave like a smart researcher. The service will receive a website/domain, investigate the site, use an LLM to infer product features, map those features to required capabilities from an internal catalogue, and return a structured JSON response with confidence, evidence, reasoning, and investigation metadata. What you will build A stateless, Dockerised HTTP service with: POST /v1/evaluate GET /health GET /version Live adaptive crawling of websites Optional login/authenticated exploration when credentials are provided Deterministic “pass-in content” mode for fixture testing Integration with an internal Catalogue API LLM-driven feature inference Deterministic feature-to-capability rollup OpenAPI 3.1 specification Contract-test harness Clean structured logging Strong secret handling: credentials must never be logged, stored, cached, or returned Key requirements The service must: Work without access to our main codebase Fetch a versioned capability catalogue at runtime Cache catalogue versions safely Never hard-code catalogue contents Never calculate readiness, serviceability, or fit scores Return useful results even when a site is gated behind login Respect crawl limits, robots.txt, time budgets, and sane rate limits Run cleanly in a container from environment configuration alone Acceptance criteria We will provide a fixture suite of frozen website snapshots and expected outputs. Your build will be accepted when it: Produces valid schema-compliant responses Is deterministic in pass-in mode Matches the deterministic capability rollup exactly Achieves the agreed inference quality bar on fixtures Handles live gated sites without failing Keeps secrets out of logs, traces, responses, and storage Ships with OpenAPI, README, Dockerfile, tests, and a one-hour handoff Preferred stack Python or TypeScript preferred. We care more about clean architecture, strong reasoning, and reliable delivery than framework preference. Ideal contractor You are strong in: Backend API development LLM application design Web crawling / browser automation Structured JSON outputs OpenAPI Docker Testing and deterministic evaluation Security hygiene around credentials and logs Experience with Playwright, Puppeteer, LangChain/LangGraph, OpenAI/Anthropic APIs, FastAPI, Node, or similar is helpful. To apply Please include: How you would architect this service Your preferred stack and why Relevant examples of LLM, crawling, or evaluation systems you have built Estimated timeline Fixed-price quote or milestone breakdown We are not looking for a low-effort scraper. We are looking for someone who can build a precise, testable, production-ready service that turns messy website evidence into structured capability requirements.