Learn how to make Large Language Models work together with other cutting-edge technologies, such as Vector Databases, Agent-Frameworks, Policy Engines or MLOps-Architectures to form innovative solutions, which deliver real value, while complying with all the possible obligations and restrictions of an enterprise environment.

Intended Audience:

• Enterprise Architects or Solution Architects who are looking for ways to securely integrate Generative AI and LLMs into productive workflows
• Data Scientists or Machine Learning Engineers, who want to learn how to design and build real-world AI solutions with Generative AI and LLMs
• (IT-)Managers or Data Protection Officers who want to gain a deeper understanding of the underlying concepts of AI Governance, including a first-hand impression of the process and implementation details

In terms of Artificial Intelligence, Generative AI, including especially Large Language Models (LLMs), can undoubtedly be considered the most widely recognized technology of the current decade (so far). However, using LLMs to build robust solutions often proves difficult for many use cases in practice or is straight up incompatible with the requirements stemming from the governance or security policy of a company.

Due to concerns of privacy and information security, the application of external Models, such as ChatGPT, via the Browser or by API oftentimes is not an option when a company’s internal information assets, personal data or other kinds of sensitive information is required as input for the LLM.

Furthermore, involving LLMs in important decision-making processes can entail imponderable risks and lead to unforeseeable outcomes, because of their non-deterministic properties and the associated deficiencies regarding the consistence of their outputs.

To round off the list of complaints, long processing times and operating cost can make it difficult and expensive to seamlessly integrate LLMs into productive workflows at scale.

Therefore, when dealing with critical processes or with sensitive data in enterprise scenarios, it is advisable to apply comprehensive patterns, designed specifically to enable organizations to leverage the power of Large Language models in a systematic, context-aware, and policy-driven manner. In this workshop, we want to show you, what these patterns look like and how they translate into actual technical solutions.

The workshop consists of two main parts:

First, we will start with a conceptual part, covering the transformative nature of Generative AI and LLMs and key concepts to follow to leverage AI-Models from a business perspective. This part will culminate in a deep dive on the game-changing aspect of LLMs on current business and their disruptive impact.

In the second part, we want to provide a structured approach to incorporating LLMs into software systems, in order to achieve controllable and consistent action execution. Then we plan get our hands dirty and walk you through the gritty details and architectural implications to consider for the design of an actual solution, followed by the specification of an exemplar technology stack for a fully-fledged AI-application, which will interconnect an LLM with a safeguarding policy engine. Finally, the resulting specification will be implemented together with you, including the use of a privacy-friendly GPT-Version on your very own computer.

It is possible to take part actively during the hands-on session, but it is also perfectly fine to watch and learn. Access to the Code Repository and detailed preparation instructions for the hands-on part will be provided upfront.

Agenda:

• 09:00 –Welcome & Introduction
• 09:15 – Start of the Conceptual Part
• 10:30-11:00 – Coffee Break
• 12:30-13:30 – Lunch Break
• 13:30 – Start of the Hands On-Session
• 15:00 -15:30 – Coffee Break
• 17:00 – Finishing the Workshop