Labra GenAI Assistant, powered by Anthropic's Claude 3.5 Sonnet and Amazon Bedrock, delivers secure, context-aware, and efficient generative AI capabilities for your Marketplace and Co-sell workflows. This section outlines the underlying technologies, processes, and safeguards that make Labra AI a reliable and secure AI assistant.
1. Foundational Model
- Anthropic Claude 3.5 Sonnet:
- A cutting-edge foundational model that empowers Labra GenAI's natural language understanding and generation.
- Enables:
- Advanced contextual understanding for personalized responses.
- Accurate and efficient processing of complex queries.
- Alignment with responsible AI principles through fine-tuned capabilities.
2. Data Flow and Integration
- Private Data Integration:
- Labra GenAI Assistant integrates with your Labra account to securely access private datasets such as:
- Private offer data.
- Co-sell opportunity details.
- CRM and AWS ACE information.
- Data Ingestion Process:
- Private data is ingested into Amazon S3 and processed into chunks for efficient retrieval.
- Amazon Bedrock Knowledge Bases utilizes vectorization to organize and index the data for semantic search.
- Labra GenAI Assistant integrates with your Labra account to securely access private datasets such as:
- Retrieval Augmented Generation (RAG):
- Labra GenAI Assistant leverages the RAG workflow:
- Retrieves relevant chunks of data using semantic search.
- Generates natural language responses by combining retrieved data with user queries.
- Example: A co-sell query retrieves customer history, partner engagement data, and deal velocity metrics for a complete response.
- Labra GenAI Assistant leverages the RAG workflow:
3. Security and Isolation
- Data Isolation:
- All customer data is stored in Amazon S3 buckets specific to your account, ensuring complete isolation.
- Vector databases (e.g., Amazon OpenSearch Serverless) ensure data remains private and secure during processing.
- Encryption:
- Data is encrypted in transit using AWS KMS and at rest using AES-256 encryption.
- Role-Based Access Control (RBAC):
- Access to private data and AI features is restricted based on team roles defined within the Labra platform.
- SOC 2 Compliance:
- Labra is SOC 2 compliant, adhering to the highest standards for security, availability, and confidentiality. Learn more at trust.labra.io.
- This ensures that all processes and systems are designed to protect customer data and maintain trust.
4. Maintenance and Monitoring
- Monitoring:
- Amazon CloudWatch tracks performance metrics and usage patterns for API calls, Slack interactions, and RAG workflows.
- AWS GuardDuty ensures continuous threat detection and alerting.
- Updates and Maintenance:
- Regular updates to foundational models (e.g., Claude 3.5 Sonnet) and guardrail configurations are managed via Amazon Bedrock.
- Seamless scaling and updates to Labra platform integrations are implemented without downtime.