AI Transforms Mining: ALS LITHOLENS™ on EKS

Revolutionizing Core Logging with AI & Cloud

The ALS GeoAnalytics LITHOLENS™ platform represents a compelling case study in applying advanced AI and cloud-native technologies to modernize a historically manual and geographically challenging industry. The key insight is the successful integration of deep learning models (RoQE Net, VeinNet, CobbleNet) with robust cloud infrastructure (Amazon EKS, S3, Lambda, RDS) to automate complex geological analysis. The innovation lies in moving from subjective, labor-intensive core logging to a data-driven, scalable, and consistent process. The use of Amazon EKS with G6 instances for GPU-accelerated ML workloads, combined with Lambda for API operations and automatic scaling down to zero, demonstrates a sophisticated approach to managing variable compute demands cost-effectively. The pre-configured AMIs significantly reduce container startup times, a critical factor for job throughput and cost optimization in ML pipelines. The business impact, evidenced by adoption across multiple mining companies and accelerated project completion, is substantial.

However, a potential limitation or area for further exploration could be the detailed discussion on the interpretability and explainability of the deep learning models. While accuracy is highlighted, understanding how these models arrive at their classifications is crucial for geologists to build trust and validate findings, especially in high-stakes decision-making for mine development. The article touches upon transparency in logging but could delve deeper into the ML model's transparency. Furthermore, while the hybrid architecture is praised for cost-effectiveness, the management complexity of a hybrid EKS/Lambda setup for a specialized domain like geology could be a consideration for smaller organizations. Comparison with other cloud providers' ML offerings or on-premises solutions isn't explicitly made, which would offer broader context. Nevertheless, for organizations within the mining sector and similar geosciences, this solution offers a clear path to enhanced efficiency, accuracy, and scalability, reducing operational costs and environmental impact.

Key Points

• ALS GeoAnalytics' LITHOLENS™ platform automates geological core logging using ML and computer vision.
• The solution leverages Amazon EKS for GPU-accelerated ML training and inference, and AWS Lambda for API operations.
• Key ML models include RoQE Net for RQD and alpha angle extraction, and VeinNet/CobbleNet for geological feature identification.
• A hybrid architecture optimizes costs by scaling EKS down to zero when idle and using Lambda for sporadic API requests.
• Pre-configured AMIs reduce container startup times, improving job throughput and efficiency.
• The platform has achieved significant business impact, including accelerated project completion and more accurate mineral detection across multiple mining companies.

---

📖 Source: How ALS GeoAnalytics LITHOLENS ™ revolutionizes core logging through machine learning with Amazon EKS