Empowering TE Connectivity's R&D with AI-Driven Knowledge Management

TE Connectivity: The Electronics Giant

• TE Connectivity is a leading global electronics manufacturing company, with operations across 3 continents and a revenue of around $16 billion.
• They produce approximately 235 billion products annually, with a team of 8-10,000 engineers responsible for designing and enhancing these products.

Challenges in Product Design and Research

• The vast volume of technical data, both structured and unstructured, spread across various repositories, makes it challenging for engineers to quickly access relevant information.
• New engineers face a steep learning curve to understand the enterprise's knowledge base, as accessing and synthesizing information can take days or even weeks.
• Ensuring secure and role-based access to sensitive data is crucial, as different teams have varying levels of data access.
• The global distribution of the engineering team adds another layer of complexity to knowledge management.

Introducing "Telme" - TE Connectivity's AI-Powered Solution

Key Features:

1. Intuitive Search and Summarization: Telme ingested 2.5 million documents and provides a user-friendly interface for engineers to search and access relevant information quickly.
2. Secure Data Access: Telme ensures role-based access to data, allowing engineers to view only the information they are authorized to access.
3. Scalability and Availability: The platform is designed to scale across multiple AWS zones, ensuring high availability and performance for 8,000-10,000 users.

Future Enhancements:

1. Expanding Data Ingestion: Telme aims to ingest up to 75 million documents to provide comprehensive coverage of TE Connectivity's knowledge base.
2. File Upload and Interaction: Allowing users to upload files and interact with the platform, even if the data is not yet ingested.
3. Generative AI Capabilities: Transforming Telme into a generative AI assistant that can handle a wider range of queries and tasks beyond the TE Connectivity domain.

Telme's Architecture and Technical Approach

• Telme leverages a serverless architecture, utilizing AWS Lambda, Step Functions, and Amazon OpenSearch for efficient data processing and storage.
• The platform ingests and preprocesses structured and unstructured data, creating embeddings and indexes for fast retrieval.
• Cross-region inference is implemented to ensure consistent and responsive user experiences, even during high traffic loads.

Key Outcomes and Learnings

• Telme achieved a 72% acceptance rate among the 8,000 engineering users, exceeding the initial 50% target.
• The platform enabled faster decision-making, secure information access, and improved productivity across TE Connectivity's teams.
• Careful data management and vectorization strategies are crucial for the success of large language model-based applications.

Conclusion

Telme, the AI-powered knowledge management platform, has transformed TE Connectivity's product design and research capabilities, empowering engineers with fast, secure, and intelligent access to a vast repository of technical knowledge.