Revolutionizing Audi's tender process with generative AI (AIM120)

Reinvent Session Recap: Streamlining Audi's Tender Process with Generative AI

Introduction

  • The presenters are Mike Wilman (Senior Data Scientist at XL2), Thomas (IT Architect at XL2), Edward (Project Lead and AI Solutions Enabler at Audi), and Simon (External PhD Student at Audi).
  • The session will cover how Audi used generative AI to streamline their tender process.

Large Language Models (LLMs) Overview

  • LLMs are large, powerful language models with capabilities like understanding and generating human-like text.
  • They are trained on vast amounts of data and use the Transformer architecture.
  • LLMs work by taking in a sequence of input tokens and generating a probability distribution of possible output tokens.

Audi's Tender Process Challenge

  • Audi's planning department is responsible for procuring specialized machinery for engine production.
  • The tender process involves describing requirements in lengthy (400+ page) documents and matching them to supplier offers.
  • This process is highly manual and labor-intensive, with over 1,000 tenders and 20,000 offers per year.

Audi's Approach

  1. Making Documents LLM-Ready: Extracting text, structure, and multimodal information from various document formats.
  2. Extracting Requirements: Using LLMs to extract a checklist of requirements from the tender documents.
  3. Matching Offers to Requirements: Matching the offer documents to the extracted requirements using semantic similarity search.
  4. Evaluation: Incorporating domain expert feedback to evaluate how well the offers meet the requirements.

Architecture Deep Dive

  • The solution uses AWS services like Batch, Step Functions, and Bedrock for scalable, managed LLM inference.
  • Key components include document pre-processing, embedding generation, and the interactive evaluation interface.

Current and Future Developments

  • Exploring fine-tuning the LLM and embedding models to better capture domain-specific knowledge.
  • Building a "golden dataset" of expert-annotated matches to enable more effective fine-tuning.

Key Takeaways

  1. Carefully plan for the costs of using generative AI.
  2. Understand the needs and processes of the domain experts.
  3. Balance human and AI interactions to create a collaborative atmosphere.

Your Digital Journey deserves a great story.

Build one with us.

Cookies Icon

These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.

If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.

Talk to us