Publications, patents, presentations and awards.

The seismic processing infrastructure Atlas Deep Geo builds on did not start with Atlas Deep Geo. It has decades of production validation behind it. Steve architected the parallel I/O and computing core of a widely-deployed commercial seismic processing system, working alongside Chuck Mosher across two organizations on the JavaSeis framework. What they built together became the foundation for multiple generations of seismic processing infrastructure used across the industry. The architectural patterns proven in that work, how data moves, how compute scales, and how processing operators are structured, are the same ones that sit at the core of Atlas Deep Geo's reconstruction stack today.

Carrying a major commercial software product through a full architectural generation without losing existing customers or breaking existing workflows is one of the hardest things a software organization can do. Steve led exactly that. He oversaw the multi-year transition of a major commercial subsurface product from on-premise deployment to cloud-native microservices and SaaS-style delivery, re-grounding legacy systems on container orchestration, distributed compute, and managed cloud services while keeping production workflows intact. The result was a fully modernized product delivered successfully to market.

Some of the most technically demanding work in the industry happens at the boundary between operator research labs and commercial software, where research-proven methods have to be turned into products that work reliably in production. Steve spent years operating at exactly that boundary. As senior engineering leader, he ran a multimillion-dollar engagement that brought a research-lab-developed seismic technology from a major operator partner into commercial deployment, and led a parallel multi-year effort to redesign traditional seismic processing workflows around cloud-native infrastructure. He served as the primary point of contact and the engineering leader responsible for turning shared research vision into commercial deliverables that shipped.

The question Atlas Deep Geo is built around, how do you make an AI-driven subsurface result honest about what it does not know, is not a new question for Steve. He is co-author of three industry publications applying machine learning to fault interpretation, with calibrated uncertainty quantification as the methodological thread across all three. The work appeared in Subsurface Insights in 2021, at SEG IMAGE in 2021, and at the SEG Annual Meeting in 2022. Getting a neural network to identify faults is one problem. Getting it to tell you how confident it is, and where it should not be trusted, is a harder one. That is the same problem Atlas Deep Geo addresses in its reconstruction work.