You want to integrate generative AI into your data pipelines but are unsure how to move from prototype to a production-ready solution. Large language models (LLMs) are powerful, but deploying them in an enterprise setting requires robust infrastructure, rigorous data management and specific engineering patterns. Azure Databricks provides a unified platform combining data engineering, MLOps and generative AI capabilities to build reliable, large-scale LLM applications. The DP-3028 training course teaches you the engineering techniques needed to put generative AI solutions into production on Azure Databricks. That is precisely the purpose of this generative AI Databricks training.
In one day at ITTA, a Microsoft Learning Partner in French-speaking Switzerland, you move from theory to practice using official Microsoft cloud labs. Your MCT-certified trainer guides you through the key steps: preparing data for RAG (Retrieval-Augmented Generation), using models via Model Serving, building processing chains with vector embeddings and evaluating the quality of generated responses. Delivered in Geneva and Lausanne in-person or as a virtual classroom, this training prepares you for the Microsoft Applied Skill assessment and helps you deploy robust generative AI solutions within your organization.
Module 1: Get started with language models in Azure Databricks
Module 2: Implement Retrieval Augmented Generation (RAG) with Azure Databricks
Module 3: Implement multi-stage reasoning in Azure Databricks
Module 4: Fine-tune language models with Azure Databricks
Module 5: Evaluate language models with Azure Databricks
Module 6: Review responsible AI principles for language models in Azure Databricks
Module 7: Implement LLMOps in Azure Databricks
Generative AI is transforming the way companies leverage their data. But moving from prototype to production remains a major challenge. Azure Databricks addresses this challenge by providing an integrated platform that covers the entire lifecycle: data preparation, model training and fine-tuning, deployment via Model Serving, and production monitoring. The lakehouse approach allows unstructured data (documents, emails, knowledge bases) that feeds RAG solutions to be stored and governed. For Swiss companies subject to strict compliance requirements, Databricks offers data and model governance mechanisms that guarantee the traceability and security of generative AI processing.
This training is aimed at data engineers, ML engineers, and developers who want to industrialize generative AI solutions on Azure Databricks. It is also suitable for data architects evaluating RAG patterns to enrich their company’s applications. An intermediate level is required: you should have a solid command of Python basics, understand fundamental machine learning concepts, and have some prior experience with Azure Databricks or a Spark environment. No prior expertise in generative AI is necessary, as the training covers foundational concepts before moving on to technical implementations.
The morning begins with a presentation of the architecture of an enterprise generative AI solution: components, data flows, and design patterns. You then discover Retrieval-Augmented Generation (RAG), the most widespread approach for combining LLMs with your organization’s proprietary data. You prepare a document corpus, generate vector embeddings, and configure a vector search index in Databricks.
The afternoon focuses on building the complete chain: you deploy a language model via Model Serving, connect the vector retriever, and implement a functional RAG application. You learn to evaluate the quality of responses with relevance and faithfulness metrics. The day concludes with governance best practices: managing model access, logging interactions, and monitoring quality in production. All exercises use official Microsoft MOC program cloud labs.
ITTA is an official Microsoft Learning Partner in French-speaking Switzerland and offers this training with the most recent Microsoft-provided materials and labs. Our MCT trainers combine data engineering expertise with practical generative AI experience in enterprise environments, enabling them to illustrate concepts with concrete use cases and share best practices observed in the field.
Small group sessions in Geneva or Lausanne, in person or virtual classroom, promote rich exchanges and allow personalized support on exercises. You can discuss your generative AI projects with the trainer and get tailored advice. You leave with the skills needed to design and deploy RAG solutions on Azure Databricks and to validate the Microsoft DP-3028 Applied Skill.
What is the difference between this training and a general LLM course?
This training focuses on production engineering: how to deploy and industrialize generative AI solutions on Azure Databricks. It does not cover in-depth transformer theory or training models from scratch, but rather the concrete patterns for putting RAG applications into production.
Is Azure OpenAI knowledge required for this training?
No, the training covers the use of models via Databricks Model Serving. Familiarity with basic LLM concepts (prompts, tokens, embeddings) is useful but not mandatory.
What is RAG and why is it so important?
Retrieval-Augmented Generation is a pattern that allows LLMs to respond by drawing on your enterprise data rather than solely on their training knowledge. It is the most widespread approach for building chatbots and AI assistants that provide reliable, contextualized responses.
Does this training cover model fine-tuning?
The training focuses primarily on the RAG pattern, which is the recommended approach for the majority of enterprise use cases. Fine-tuning may be mentioned as a complement but is not the core of the program.
Is the DP-3028 Applied Skill complementary to other certifications?
Yes, it combines naturally with the DP-203 certification (Data Engineering) and the DP-3027 Applied Skill (Databricks Data Engineering). Together, they cover a comprehensive range of skills on the Databricks platform.
Is the data used in the labs kept confidential?
The labs use Microsoft-prepared datasets for training purposes. You do not manipulate your own company data during exercises, which eliminates any confidentiality risk.
Can these skills be used with platforms other than Databricks?
The concepts of RAG, embeddings, and LLM evaluation are transferable to other platforms. However, specific technical implementations (Model Serving, Vector Search) are specific to the Databricks ecosystem.
Monday to Friday
8:30 AM to 6:00 PM
Tel. 058 307 73 00
ITTA
Route des jeunes 35
1227 Carouge, Suisse
Monday to Friday, from 8:30 am to 06:00 pm.
