04-23, 14:45–15:30 (Europe/Berlin), A1
Last year, dm-drogeriemarkt was among the first big German companies launching a tool for the coworkers to be able to unlock the power of LLMs in a secure setting. At the beginning, dmGPT was only a user interface pointing to a private instance of a foundation Model.
Listening to the needs of our colleagues, we quickly learned that this “naked” model – a super powerful NLP Model that can help them processing text - is not really what they needed: they needed a trustworthy, knowledge-rich assistant to help them accomplish their daily tasks.
In our journey towards this goal, we used python to shift the LLM’s role in dmGPT: from being the motor and only source of answers to being a translator between the user’s input in natural language and multiple software systems, the steering wheel that helps humans drive the flow.
Today, dmGPT is not only a statistical parrot anymore, now it is an open platform powered by internal knowledge.
In this talk we want to share with you the learnings and insights we gained while designing and implementing the new dmGPT.
One of the biggest challenges of working in such a large organization like dm is finding the information you need to accomplish your tasks: distributed organization units, multiple knowledge sources, and different tools make it very challenging to know where to find information whose location you don’t know. Most of the times, the best way to find something out is to ping a more experienced colleague and ask them. But what if you could ping your AI-Powered copilot and find out? Not only that… What if it also helped you create content for your specific product without you telling it everything about the product? What if it was able to help you write code using internal tools? What if it could help you have an insight of your internal data?
After its first steps in summer 2023, our vision for dmGPT quickly developed to it becoming a truly helpful assistant for every coworker of dm. Since then, we have contributed to the design and implementation of an LLM-powered platform that aims to achieve this goal. To come a step closer, we had to rethink the role of the LLM, picturing it as a translator between natural languages and software systems and back. Now, it helps us map an instruction in natural language to a set of tools needed to accomplish the given task and construct a coherent answer based on the provided data.
In the design we had to face multiple challenging questions, such as:
- How to connect multiple, heterogenic data sources?
- How to pick an LLM for a given task?
- Which LLM do we support?
- How do we build a user friendly, dynamic and configurable user interface?
- How to measure the system’s quality?
In this talk we would like to provide a technical insight to our journey, discussing architectural decisions as well as implementation dilemmas, and engage in a discussion with the community about the steps to come.
Intermediate
Expected audience expertise: Python –Novice
Abstract as a tweet (X) or toot (Mastodon) –Learn how dm-drogeriemarkt put LLMs in production and implemented a day-to-day assistant for everyone.
Half colombian, half spanish half german data person with particular interest in natural languages. Translator & Computer Scientist by training, pythonista by heart.
I'm a data person with a focus on machine learning and data science but also business background. Working for a few years now in the data world - especially with Python - I'm now the product lead for a team called Customer genAI Incubator at dmTECH which is the IT subsidiary of dm-drogerie markt.