GENAI-Powered chatbot to increase team efficiency
summary
Audi deploys a generative-AI chatbot that indexes cleaned, KMS-encrypted Confluence PDFs into OpenSearch, converts text to embeddings with Longchain in AWS Lambda, and serves answers via a React frontend behind Cognito SSO; the stack trials Llama and Falcon on SageMaker then switches to Amazon Bedrock for managed models, guardrails and crawlers. It supports 400 accounts and 4,000 developers, exposes feedback via DynamoDB thumbs up/down, uses hexagonal architecture for model swaps, and yields measurable reduced search time and lower manual ticket volume.
AI-written summary · about
chapters (6)
- 0:00Introduction and context
- 10:53Generative AI overview
- 21:12Chatbot architecture details
- 27:58Demo and challenges
- 30:28Responsible AI and improvements
- 35:04Key takeaways and conclusion
related talks
47mOutsmarting the Genie: Smart Architecture Decisions for GenAI Apps
20mExplore what’s possible with new AWS generative AI services
22mLearning to Apply Generative AI at Enterprise Level
38mCraft your own Generative AI Chatbot with Amazon Bedrock
28mUnlocking the power of GenAI
33mOperationalizing GenAI Applications with DevOps Practices
51mTransforming Banking with Advanced AI: GenAI, RAG, and Multi-Agent Systems
29mCustomizing genAI applications for your business