What has been your approach to effectively integrating AI/ML models and tools into your existing IT infrastructure and applications?

We engage in extensive modeling, particularly for environmental impact assessments related to new developments. Our models predict outcomes such as runoff and environmental changes due to construction. While our applications haven't fully integrated AI tools, we rely on precise data inputs to ensure accuracy. This is crucial, especially when predicting the impact of significant weather events like hurricanes. We rebuild and recalibrate models when updating application versions to ensure consistent and accurate data processing. In the scientific realm, there's a lag in AI integration, but it's essential to maintain accuracy and reliability in our models. The stakes are high, so we need to ensure that applications perform as expected before integrating AI tools.

We're currently exploring the best methods for integrating AI/ML into our infrastructure. One promising approach is to create a centralized institutional environment—a platform that houses all necessary data. AI requires not only a large volume of data but also high-quality data. By centralizing data, possibly using data lakes or lake houses, we can manage access to the data and integrate various AI/ML tools and models, whether developed internally or purchased. This platform would ensure proper security controls, governance, and data management, allowing us to protect data and intellectual property effectively. It's crucial to manage data access carefully, ensuring that only authorized individuals can access sensitive data. To support this, we recognize the need to enhance our data governance and management practices. It's a roadmap that requires gradual development and refinement.