ThirdBrAIn.tech

From PoC to Production AI at Enterprise Scale | WSO2Con Barcelona 2025

2 min read

From PoC to Production AI at Enterprise Scale | WSO2Con Barcelona 2025

AI Summary

Summary of AI Application Development

Speaker: Malit Jing, VP of Research and AI at WSO2

Overview

  • Discussed transitioning AI applications to production.
  • Shared experiences from past AI development practices.

Key Highlights

  1. Historical Context
    • Early AI predictions using Ballerina programming language.
    • Emphasis on extensive data collection, preprocessing, and model training.
    • Shift from specialized model training to leveraging large language models (LLMs).
  2. Current Development Practices
    • Use of LLMs for integration without the need for extensive preprocessing.
    • Importance of integrating AI components into larger applications.
    • Common challenges faced when deploying AI projects (74% failure rate).
  3. Integration Challenges
    • Issues with evaluation, security, governance, and monitoring in production environments.
    • Necessity of new capabilities and abstractions in software engineering to address challenges.
  4. Development Tools
    • Choice of tools depends on developer background (low code vs. pro code).
    • Popularity of low-code development tools due to standardized patterns for quick application building.
  5. Case Study: WSO2 Mobile App
    • Integration of AI capabilities to enhance user experience in the WSO2 conference mobile app.
    • Features included chat assistance, session recommendations, and attendee connections.
    • Challenges with user consent for accessing additional data sources like LinkedIn.
  6. Future Directions
    • Ongoing development of agent concepts to enhance user interactions and experiences.
    • Introducing capabilities for monitoring and evaluation of AI systems.
    • Expansion of AI gateway functionalities to manage AI traffic and enhance security.

Conclusion

  • Bridging the gap from prototype to production involves navigating complexities in integration, governance, and security. Future developments will continue to enhance the capabilities of AI applications, aiming to reduce the current high failure rates in production scenarios.

Graph View

Backlinks