AWS re:Invent 2024 -From AI prototype to enterprise-grade gen AI app with Informatica & AWS (AIM245)

The video summarizes Informatica's approach to enabling Enterprise-grade Generative AI (Gen-AI) applications. Here are the key takeaways:

Requirements for Enterprise-grade Gen-AI

  1. Groundedness: Avoiding hallucination and inaccurate responses by grounding the models in Enterprise data.
  2. Contextualization: Ensuring responses are relevant to the specific Enterprise context and terminology.
  3. Quality: Ensuring the input data has the required level of quality to generate accurate and reliable responses.
  4. Ease of Development and Deployment: Enabling rapid development and deployment of Gen-AI applications using low-code/no-code approaches.
  5. Governance and Security: Ensuring transparency, data traceability, data access control, and cost management for Gen-AI applications.

Informatica's Approach

  1. Scalable Data Transformation and Integration: Leveraging Informatica's capabilities for scalable data integration from diverse sources.
  2. Data Quality and Observability: Ensuring high-quality, accurate, and unbiased data feeds for Gen-AI applications.
  3. Explainability and Traceability: Providing lineage and explainability for Gen-AI responses using metadata intelligence.
  4. Semantic Intelligence: Guiding the language models with domain-specific semantic intelligence.
  5. Sensitive Data Handling: Addressing sensitive data issues and enforcing access management policies.
  6. Unified Master Repositories: Creating a consistent, trustworthy, and comprehensive foundation for Gen-AI responses.
  7. Simplification of Gen-AI Development: Providing a no-code/low-code orchestration platform with cross-language model support.

Evolution from Retrieval-Augmented Generation (RAG) to Agents

  • RAG frameworks have challenges in production, such as hallucination, data freshness, and compliance requirements.
  • Agents are more autonomous, task-oriented, and better integrated into Enterprise systems, leading to higher success in production.

Enterprise-grade Gen-AI Architecture

  1. Agent Subprocess: The core of the architecture, responsible for understanding user intent, planning, orchestrating, and summarizing responses.
  2. Planner: The "brain" of the agent, determining the optimal plan to retrieve and integrate data from various sources.
  3. Orchestrator: Executing the plan by calling the appropriate data executors and managing the state.
  4. Executors: Specialized components for retrieving data from specific data sources, leveraging Informatica's pre-built accelerators.
  5. Security and Governance Proxy: Enforcing data access policies and regulations based on user identity and context.
  6. Summarizer: Distilling the integrated data into a coherent, explainable, and Enterprise-relevant response.
  7. Front-end Layer: Handling user authentication, authorization, and rate limiting for cost control.

The video also showcases a live demo of the Enterprise-grade Gen-AI application, highlighting the use of metadata, data quality, and cross-system integration to provide contextual and explainable responses to a supply chain analyst.

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