Description:
Build a production AI platform that creates and updates training content for regulated sectors (health, care, education). You’ll own end-to-end delivery: intake-gated chat, audited RAG with citations, agent pipelines (slides/quiz/SCORM/blog), allowlisted web connectors, queues/schedulers, multi-tenant data, and audit/compliance.
What you'll do:
- Ship the full stack: Next.js UI → API → DB → AI layer.
- Implement RAG with embeddings, chunking, metadata filters, and citation enforcement (“no source, no answer”).
- Build deterministic agents (script → slides, quiz, SCORM, blog) and export pipelines.
- Create allowlisted web connectors (e.g., Ofsted/CQC), with robots.txt compliance and change-detection (hash+diff).
- Set up background jobs/queues/schedulers (Redis/Celery/RQ/SQS).
- Implement multi-tenant models, RBAC, PII handling, audit logs, basic cost/rate limits.
- Collaborate on simple, reliable UX (intake-before-chat; “Send to Agents”).
Must-have experience:
- 4+ years full-stack product engineering (React/Next.js + Python/FastAPI or Node/NestJS)
- Shipped production LLM/RAG features (not just experiments), with citations
- Postgres (+ pgvector or Pinecone/Weaviate), schema design, migrations
- Redis (queues/schedulers), S3/GCS storage, exports (docx/pptx/pdf/SCORM)
- OpenAI Assistants/Responses API with tool/function calling; prompt templating
- Security/compliance mindset: RBAC, PII, observability, audit trail.