Welcome to the Chatbot Essentials channel. This folder contains the meeting recordings of every session held so far (10 and counting) and the resources for the series. The sessions take place every second Thursday and are free to attend — register here.
Table of Contents
Registration
🔐 Register for the Webinar Series
To register for the Chatbot Essentials webinar series, please visit: Free AI Webinar Series - Master AI Chat & LLMs
Follow the link to complete your registration and get access to all upcoming sessions.
Planned Sessions
📅 Webinars are held every second Thursday at 17:00 - 18:45 Central European Time (CET).
Mark your calendar and join us for in-depth discussions on chatbot essentials, best practices, and live Q&A sessions.
Webinar Structure: Each session is divided into a theory part and a hands-on exercise part. t2b provides all necessary tools and infrastructure, so you can focus entirely on learning and practicing.
Meeting Platform: All webinars take place online via Microsoft Teams. You will receive meeting links and join details upon registration.
Calendar
| Date | Description |
|---|---|
| Thursday, May 21, 2026 — ▶️ Play | Introduction to LLMs: What is a Large Language Model; meaning of "G" (Generative), "P" (Pre-trained), "T" (Transformer); positive and negative consequences of each component; counter-measures to avoid potential mistakes; hard-core limitations |
| Thursday, June 4, 2026 — ▶️ Play | Prompt Engineering Exercises and Best Practices - Hands-on exercises with different prompting techniques; testing and iterating prompts; prompt chaining; few-shot learning; zero-shot learning; practical strategies for improving model outputs |
| Thursday, June 18, 2026 — ▶️ Play | What is a Chatbot? What's the difference between an LLM model and a Chatbot? A Chatbot's architecture and the most important data structure. Influencing the LLM answers with system prompt engineering. Notions of Chatbot memory. Notions of RAG. Notions of tool calling. |
| Thursday, July 2, 2026 — ▶️ Play | Tool Calling Part 1: Fundamentals - Tool calling basics; tool orchestration and planning; data dependencies and time dependencies; dependency injection patterns; how small LLMs can orchestrate complex tasks |
| Thursday, July 16, 2026 | Summer holiday break |
| Thursday, July 30, 2026 | Summer holiday break |
| Thursday, August 13, 2026 | Summer holiday break |
| Thursday, August 27, 2026 — ▶️ Play | Tool Calling Part 2: Advanced Calling - Advanced calling techniques; multiple tool calling; data dependency injection in tool parameters; notions of planning and agents orchestration |
| Thursday, September 10, 2026 — ▶️ Play | RAG Part 1: Embeddings & Search - Embeddings spaces, similarity search, quasi-orthogonality; simple RAG fundamentals; vector databases and retrieval strategies |
| Thursday, September 24, 2026 | Holiday Break |
| Thursday, October 8, 2026 — ▶️ Play | RAG Part 2: we build a memory subsystem in a chatbot! Then we see cognition management; memory optimization and pruning; recall systems; quality systems for source data improvement |
| Thursday, October 22, 2026 — ▶️ Play | RAG Part 3: Building a Chatbot Memory Subsystem |
| Thursday, November 5, 2026 — ▶️ Play | RAG Part 4: Advanced Techniques & Knowledge Graphs - A live case study first: the memory consolidator had merely concatenated two facts about Bob the cat instead of merging them, and the fix was pure prompt engineering — a permissive "can be merged" that had to become a "must", plus one worked example, because a small model on a small machine is led by hand and a good example convinces better than a good instruction. Then ingestion in depth: extraction is the hard part (PDF is a typographical binary format, so Docling plus deep-learning OCR), followed by chunking along the document's logical structure, chapter summaries consolidated into a document summary, curated categories instead of free-form keywords, and embeddings at ~4 KB each — which is why the database alone wants tens of gigabytes of working cache. Then knowledge-graph RAG: human and model co-design an ontology, entity, relationship and attribute extraction run massively in parallel (batching raises throughput and the chip temperature), and a consolidation stage merges duplicates such as "bob" and "Bob the Cat". Finally, contradictory sources: every fact keeps its provenance, a named data steward can overrule a stale document with a dated authoritative statement, sources are ranked by trust, and the answer format is configurable per role — from a plain verdict to the full trail of deviations |
| Thursday, November 19, 2026 — ▶️ Play | RAG Part 5: Recap & RAG Optimization in Production - Recap of why RAG is needed: a frozen pre-trained model can learn and remember nothing, so RAG augments each question with context retrieved from the organization's own documents, repositories and knowledge graphs, grounding the answer in knowledge the model never saw. A real-world case: a tender with a hard 15-second end-to-end response deadline forced us to change the RAG algorithm itself while we were still 10 seconds away from the target. Why RAG is a never-ending story - optimization is content-specific, so a single engine has to support website crawling, single-point-of-truth knowledge bases and everything in between at once; retrieval from internal and external backends is left for the tool-calling sessions |
| Thursday, December 3, 2026 | LLM Architecture & Model Selection - Understanding different model families and architectures; comparing performance characteristics; selecting models for specific tasks; quantization and parameter efficiency; multi-model orchestration strategies |
| Thursday, December 17, 2026 | Fine-tuning & Model Optimization - Parameter-efficient methods: LoRA, adapters, and prefix tuning; quantization and pruning techniques; evaluation metrics and benchmarking; deployment optimization; production considerations |
| Thursday, December 31, 2026 | AI Development Team Workflows - Prompt engineering, agent governance, and optimization; mixed human-AI teams; achieving advanced automation; preventing loss of control in complex use cases; practical tips and tricks for production deployments |
| Thursday, January 14, 2027 | Thalabus Deep Dive - Thalabus architecture and REST APIs; Python SDK; embedding Thalabus modules in your application; extending and customizing Thalabus; configuring business use cases; defining custom agents, prompts, and tools; connecting databases and frontends to Thalabus |
| Thursday, January 28, 2027 | Practical AI Services & Integration - Using ready-built Thalabus services: document translation, transcription, question answering, natural language to Cypher/SQL queries, speech-to-text; accessing Thalabus RAG from external applications; Thalabus query language |
| Thursday, February 11, 2027 | Advanced Use Cases & Real-World Applications - Complex reasoning systems; deep reasoning templates; connecting enterprise databases and backends; case studies and practical implementations; Q&A and community showcase |
| Thursday, February 25, 2027 | Multi-modal AI & Vision Systems - Image understanding and generation; vision-language models; audio processing and speech synthesis; multi-modal reasoning; building multi-modal applications with thalabus |
| Thursday, March 11, 2027 | Scaling & Deployment - Infrastructure for production AI; load balancing and fault tolerance; monitoring and observability; cost optimization; distributed inference and service orchestration |
📹 Meeting Recordings
- Session 1: Introduction to LLMs — ▶️ Play
- Session 2: Prompt Engineering Exercises and Best Practices — ▶️ Play
- Session 3: What is a Chatbot? — ▶️ Play
- Session 4: Tool Calling Part 1 — ▶️ Play
- Session 5: Tool Calling Part 2 — ▶️ Play
- Session 6: RAG Part 1: Embeddings & Search — ▶️ Play
- Session 7: RAG Part 2 — Chatbot Memory & Cognition — ▶️ Play
- Session 8: RAG Part 3 — Building a Chatbot Memory Subsystem — ▶️ Play
- Session 9: RAG Part 4 — Advanced Techniques & Knowledge Graphs — ▶️ Play
- Session 10: RAG Part 5 — Recap & RAG Optimization in Production — ▶️ Play