Gen AI Leadership Forum · Engineering Bootcamp
Building Production-Ready Knowledge Systems
A one-day AI course in production-grade GraphRAG — semantic search fused with knowledge-graph reasoning, a full 6-step ingestion pipeline and a 4-level fallback chain, built hands-on in TypeScript with zero notebooks.
0
critical RAG failure modes directly solved
0
step ingestion pipeline built end-to-end
0
production codebases to take home
0
learning blocks, zero skill gaps
[ Core Stack ]
[ Five Queries Flat Retrieval Misses ]
This AI course starts where pure RAG stops — it retrieves similar text, but the questions that matter in production need logic: multi-hop, numerical, temporal and relational.
>Who is the CEO of the company that acquired John’s startup?
✗ Vector search only
Returns the chunks that mention John. The acquisition and the CEO sit in other documents, so the answer never assembles.
✓ GraphRAG
Traverses FOUNDED → ACQUIRED → WORKS_AT across typed edges in Neo4j and returns the person at the end of the chain.
>Which companies reported revenue growth above 20% in Q3?
✗ Vector search only
Similar wording scores high, but the number itself is never compared, so the results cannot be trusted.
✓ GraphRAG
MONEY and DATE entities from BERT NER become node properties that a Cypher query filters exactly.
>Which deadlines moved after the Q3 review?
✗ Vector search only
Deadline and trend queries come back without the ordering and precision a business decision needs.
✓ GraphRAG
Dated entities and typed edges let the query planner order events and compare them before and after a point in time.
>Summarise everything we know about this supplier.
✗ Vector search only
Answers that need four or more documents come back incomplete and cannot be synthesised reliably.
✓ GraphRAG
Mutual indexing links every entity to the chunks that mention it through MENTIONED_IN edges, so the answer is assembled across documents.
>Which products depend on this failing service?
✗ Vector search only
A flat vector store has no dependency edges to follow, so nothing useful comes back.
✓ GraphRAG
Walks the relationship edges from the service out to every product that depends on it.
[ Who Should Attend ]
Software Engineers
Adding retrieval to production systems that must answer multi-hop, numerical and temporal questions, not just return similar text.
AI Engineers
Building an AI chatbot that answers from company data, with a fallback chain that keeps answering when a service is down.
Tech Leads
Choosing between pure RAG, KAG and GraphRAG, and routing each query to the retrieval it actually needs.
Data & ML Engineers
Moving from vector search to knowledge graphs, with entities and relationships extracted at ingestion.
Solution Architects
Designing hybrid retrieval on Milvus and Neo4j, from the ingestion pipeline to result fusion.
Backend & Platform Teams
Running the retrieval stack on Docker Compose and TypeScript services the team can deploy and maintain.
[ The 6-Step Ingestion Pipeline ]
Every step of the AI course is implemented in TypeScript, and every choice — chunk strategy, HNSW parameters, fusion weights — comes with the reasoning behind it.
Split each document at sentence boundaries into chunks that keep context across the seams.
Generate embeddings for every chunk and store them in Milvus with HNSW indexing.
Run BERT NER over every chunk to identify the entities the graph is built from.
Use an LLM to identify typed relationship edges between the extracted entities.
MERGE entities and relationships into Neo4j as typed nodes, edges and Cypher properties.
Link graph entities and text chunks in both directions so one query can use both.
[ Intelligent Query Processing ]
In this AI course you build a query planner that classifies each question — semantic, structural or hybrid — runs vector and graph search in parallel when both are needed, and fuses the results 60/40.
[ AI Course Agenda & Curriculum ]
The AI course starts from the real limits of RAG, so every KAG and GraphRAG concept arrives when it is needed.
Understand why RAG fails on complex queries and how knowledge graphs solve logical, numerical and relational gaps.
Implement the full 6-step ingestion pipeline — from raw document to a queryable, mutually indexed knowledge graph in TypeScript.
Build a query planner that routes to the right retrieval strategy and fuses vector and graph results.
Build three real-world systems — academic research, legal case law and customer support — each powered by hybrid GraphRAG.
Optimise, harden and monitor the GraphRAG system with resilience patterns, tuning and production-grade observability.
Leave with a decision framework for when to use each approach and a set of next projects to consolidate the skills.
[ Three Codebases From the AI Course ]
Each lab in this AI course runs about 40 minutes and ends with working hybrid GraphRAG code — not an exercise, a codebase adaptable to real enterprise problems.
Academic research
Legal and case law
Customer support
[ The 4-Level Fallback Chain ]
Circuit breakers, timeouts, Redis caching and p95 monitoring are part of the AI course from the first block, so the system you build degrades gracefully instead of going silent.
Vector and graph in parallel via Promise.all()
Milvus HNSW search only
Subgraph cache lookup, near-instant when warm
Graceful degradation that always returns something useful
[ Your Instructor ]
AI Course Instructor
Sr. Software Engineer, ENTERSOFTONE
Alexandros is a Senior Software Engineer with 15 years of versatile experience in full-stack development. He specializes in C#/.NET, React.js, Python, WebGL, and AI development, including agent-based systems, RAG workflows, and Mastra AI SDK integration.
He also brings strong DevOps expertise with Azure, enabling scalable and secure deployments. A creative problem solver, he combines deep technical knowledge with intuitive UX/UI design, consistently staying ahead of industry trends to deliver modern, engaging applications.
[ Questions Engineers Ask ]
No. Every lab is written in TypeScript on Node.js — deployable code, no Python notebooks.
Yes. Unlike most AI education, the day runs from the Docker Compose setup at 09:00 to three working GraphRAG codebases in TypeScript, with hands-on labs along the way.
Intermediate to advanced. The day starts from the limits of RAG and moves quickly into graph construction, query routing and production hardening.
Three working codebases — a research assistant, a legal document search and an intelligent support system — adaptable to your own data.
A laptop with Docker Compose and an OpenAI API key. Milvus and Neo4j start from Docker Compose during registration at 09:00, where the key is validated.
October 21, 2026, at Telekom Academy in Athens, from registration at 09:00 to the wrap-up at 18:30, as part of the Gen AI Leadership Forum.
Reserve your seat on the Gen AI Leadership Forum tickets page.
[ More Forum Bootcamps ]
Engineering · Oct 22
One day building agents that run on their own — a support agent, tiered memory and a multi-agent research system you keep.
Learn More →Product & Operations · Oct 22
One day that gives everyone one to three hours back, every day — no code, just the work rebuilt around AI.
Learn More →Leadership · Oct 22
One day for the people who set direction — seven modules across four strategic phases, no code.
Learn More →Eight hours, five blocks and three GraphRAG codebases in one AI course — October 21, 2026, at Telekom Academy in Athens.