Building a RAG System for an Educational Platform
How Kilele Archade grounds its AI tutor in real curriculum documents: ingestion, chunking, pgvector retrieval and provider fallback, not model memory.
Blog / AI Engineering
AI Engineering
Retrieval-augmented generation, pgvector retrieval, LLM integration and provider fallback — how the AI in Kilele Archade actually works.
Most AI features fail in production for boring reasons: the model answers from memory, nobody can tell where an answer came from, and one vendor outage takes the product down. Every AI system I have built starts from those three problems rather than from a model.
The pattern that has worked for me is retrieval-augmented generation grounded in material the product already owns — curriculum documents in Kilele Archade's case — with embeddings stored in PostgreSQL next to the relational rows they describe, a fallback provider layer behind the primary one, and a fixed question set so a retrieval change is measured instead of argued. The Lab holds the benchmarks that came out of that work.
Writing on ai engineering (2)
Designing AI-Powered Learning Systems
What building Kilele Archade taught me about where AI genuinely helps learners — and where plain software engineering does the job better.
Built in practice
Hire for this
Experiments in the lab
RAG Strategy Bench
Side-by-side evaluation of chunk sizes, embedding models and retrieval strategies on the same document set — the harness that picked Kilele Archade's retrieval settings.
AI Provider Fallback Harness
A test rig that simulates provider failures — timeouts, rate limits, malformed responses — to verify transparent failover across LLM backends.