How E.On productionizes its AI model & Implementation of Secure Generative AI.

Stop just prompt engineering. E.On reveals how they build production-grade LLM guards to secure their AI and solve core business challenges.

#1about 2 minutes

Defining the roles of data scientists and MLOps engineers

MLOps engineers build secure, automated infrastructure to help data scientists move models from proof-of-concept to production in weeks instead of months.

#2about 2 minutes

Implementing the data as code concept for ML

E.On uses a "data as code" approach with versioned Python libraries to provide data scientists with abstracted, quality-checked data frames.

#3about 3 minutes

Using LLMs to discover datasets and manage metadata

LLMs can query enterprise metadata, helping users find relevant data sources and even generate SQL queries through natural language conversations.

#4about 1 minute

Powering website search with generative AI

E.On is replacing traditional keyword search on its website with a GPT-powered system that provides crisp, human-like answers based on a broad knowledge base.

#5about 2 minutes

Using LLMs to understand and navigate codebases

Developers can use a combination of LangChain, vector databases, and OpenAI to ask natural language questions about a large codebase for faster debugging.

#6about 1 minute

Architecture for analyzing contact center call recordings

An architecture using speech-to-text and Azure OpenAI analyzes customer call recordings to extract sentiment, verify GDPR consent, and uncover business insights.

#7about 2 minutes

Automating email processing with intent classification

To manage millions of customer emails, an automated system uses OpenAI for intent classification to suggest accurate and fast responses for call center agents.

#8about 3 minutes

Securely connecting generative AI to enterprise data

An architecture combining a knowledge base, cognitive search, and embeddings allows GPT to securely answer questions using private enterprise data, protected by LLM guards.

#9about 1 minute