Seek and
we shall find.
ORNL wrote it all down. All of it. We built them a way to find any of it.
Searching shouldn't be harder than the science.
As the nation's premier laboratory focused on everything from clean energy to climate change, the volume of data being sifted, sorted, and sought after at ORNL was immense. Things were getting missed.
The existing toolset was the trifecta of technical debt: inaccurate results, hard to update, poor user experience. For an organization this data-intensive, finding the policy on how to get reimbursed for your superconducting supercomputer shouldn't be as hard as programming the thing.
Three layers of custom infrastructure.
Rather than just plugging in a new tool, we engineered a back-end designed for flexibility and precision — built around how ORNL's teams actually use search.
Core engine
Elasticsearch implementation tuned to handle massive scientific datasets with high-speed retrieval across the lab's full corpus.
ElasticsearchCustom API layer
A robust Python-based backend bridges the gap between ORNL's complex data structures and a custom front-end UI built for the way researchers actually work.
Python · RESTInfrastructure & DevOps
Deployment tooling, programmatic testing, and ownership of the development workflow — so the system is stable, observable, and production-ready.
CI/CD · TestingHalfway through, the hypothesis changed. So did our role.
The best experiments start with a hypothesis, test against it, and adapt when reality pushes back. When the original SOW with Elastic shifted, our team shifted with it — moving from consultants to embedded partners.
During initial data ingestion, we hit significant compatibility walls between Elastic's connectors and ORNL's legacy SharePoint and ServiceNow systems. Our team spent weeks debugging and tailoring those connectors. The result has persisted through every major version upgrade since.