Description
Summary The product: We're building an AI-powered exam prep platform for a professional licensing exam (trade/certification space). It replaces static question banks with a diagnostic-first, adaptive system: students take a diagnostic, get routed to a personalized study path, and work through a multi-agent tutoring system until they hit a measured "ready" threshold. This is a live commercial product with a defined architecture already specced out — you're not starting from a blank whiteboard, but you are doing the real build. Why this is phased, not a single sprint The full scope below is substantial — a knowledge base pipeline, a multi-agent tutoring engine, payments, CRM integration, and QA automation is a multi-month build done properly, regardless of how much AI-assisted coding speeds up implementation. We're hiring for Phase 1 (MVP) with the expectation of continuing into Phase 2 if it's a good fit. We're not looking for someone to rush the whole list at once — we're looking for someone who ships a working core fast and doesn't cut corners on the parts that are hard to fix later (data architecture, cost control, security). Phase 1 — MVP scope: 1. Knowledge base pipeline: ingest source material and generate an abstraction layer (not direct rephrasing) that downstream content generation draws from — plus a way to add new source content over time without a rebuild. 2. Diagnostic engine: assembles a scored diagnostic, classifies per-topic strength/weakness, and routes the student to a tier/path. 3. Multi-agent tutoring system: several narrow, purpose-built AI agents (not one general chatbot) — e.g., a structured-topic tutor, an adaptive practice/reinforcement agent, and a live Q&A tutor. Prior experience with multi-agent orchestration and RAG is required, not just "calling an LLM API." 4. User accounts + payment: signup/auth, Stripe integration, tier/access gating. 5. Core UX/UI: functional, clean, mobile-friendly — polish comes later, usability doesn't. 6. Token/cost optimization: pre-generate and cache wherever quality allows; reserve live model calls for the moments that actually need them. You should be able to speak concretely to how you'd control inference cost at scale, not just "use a cheaper model." Phase 2 (future scope, not this contract) CRM integration (GoHighLevel), AI-assisted QA/QC tooling, documentation, and ongoing maintenance/feature expansion. We'll discuss this once Phase 1 ships and we've seen how we work together — we are not committing to this scope or a retainer in this posting. Must-have experience: • Shipped production systems using LLM APIs (Anthropic or OpenAI) — specifically multi-agent or multi-step orchestration, not single-prompt chatbots • RAG / knowledge base architecture • Stripe integration • Comfortable working from a written spec and flagging gaps or risks in it, not just building literally what's written • Can talk specifically about how you'd control LLM inference cost at scale Nice to have • Experience with Lovable or replit, or willingness to work in it • GoHighLevel API experience • EdTech or assessment/diagnostic-scoring background To apply, include: • 2–3 examples of relevant work (ideally something with LLM agents or RAG in production, not just a demo) • Your fixed price or rate for Phase 1 specifically, based on the scope above • Realistic timeline estimate for Phase 1 • One thing in the spec above you'd push back on or do differently, and why