We’ve launched the new AI Engineering Course for SurrealDB University, a free, hands-on course covering the core techniques behind modern AI applications: vector search, RAG, hybrid search, agent memory, text-to-SurrealQL, context-layer architecture, chunking strategies and graph RAG.
AI engineering is increasingly about more than calling an LLM API. As applications move from simple prompts to AI agents that need to search, reason over and remember real-world information, engineers need to solve a new set of problems. Some of these are:
How do you retrieve the right context?
How do you combine semantic and keyword search?
How should an agent remember information?
How do you give multiple agents access to data safely?
That’s what our newest SurrealDB University course is designed to teach.
Learn the foundations of AI engineering
SurrealDB for AI Engineers is an eleven-lesson course that takes you from understanding how LLMs use context to building the retrieval and memory layer behind more sophisticated AI systems.
The course covers:
AI and context engineering fundamentals: tokens, context windows, hallucinations, prompt engineering and context engineering.
Vector embeddings and vector search: semantic retrieval, similarity search and HNSW indexes.
Full-text search and BM25: finding exact terminology, identifiers and keywords that semantic search can miss.
RAG knowledge bases: combining semantic retrieval, structured filters and relationships to give LLMs better context.
Hybrid search and reranking: combining vector search and BM25 using Reciprocal Rank Fusion.
Agent memory: recalling information based on similarity, recency and importance, while handling forgetting and updating memories over time.
Text-to-SurrealQL: allowing an AI agent to generate queries against your database while keeping the schema and guardrails under your control.
Context-layer architecture: designing systems where multiple AI agents can securely access information spread across databases, documents, SaaS applications and data lakes.
Chunking strategies: splitting documents into retrieval-friendly chunks using size, overlap, structure and semantics to preserve context without introducing unnecessary noise.
Evaluating retrieval quality: measuring whether retrieval returns the right context using relevance, recall and ranking metrics, and identifying where retrieval pipelines are failing.
Graph RAG beyond one hop: traversing multi-hop relationships to uncover connected entities, dependencies and context that cannot be found through similarity search alone.
Built for engineers
The course is practical and designed for developers who want to learn AI engineering by building. You don't need previous SurrealDB experience. You don't need a cloud account or even an embedding API to get started.
From the second lesson onwards, you'll run SurrealDB locally and work through small schemas, datasets and queries. The examples deliberately keep the underlying mechanics visible so you can understand how retrieval works before replacing the examples with production embedding models and your own data.
Each lesson also stands on its own, so if you're already familiar with the basics you can jump directly into RAG, hybrid search, agent memory or context-layer architecture.
Go from RAG to agent memory and context layers
The lessons build progressively:
Vector search + full-text search → RAG → hybrid search → agent memory → agent-generated queries → context layer → chunking strategies → evaluating retrieval quality → Graph RAG
This mirrors how many AI applications evolve.
A simple RAG application might begin by retrieving documents based on semantic similarity. As the system grows, you may need exact keyword retrieval, structured filters, relationships between entities, persistent memory, permissions and access to multiple data sources. At that point, retrieval becomes a broader context engineering problem.
SurrealDB's multi-model architecture lets you work with structured data, documents, graphs, vectors and full-text search together. The course shows how these capabilities can be combined rather than treated as separate infrastructure problems.
Start learning AI engineering
With this, SurrealDB University now has six courses in total: four ways to learn SurrealQL depending on your time and learning style, schema internals and migrations, and now AI engineering.
If you're learning how to build RAG applications, AI agents or agent memory systems - or you're trying to understand what sits between an LLM and your organisation's data - SurrealDB for AI Engineers gives you a practical place to start.
