Oil and gas companies want to use AI on their most sensitive technical data, but many are wary of sending it to someone else’s cloud. GeoComputing Group, a Houston provider of high-performance computing for the energy sector, says it has an answer. On October 13 the company announced AViR, a language-model service built for oil and gas technical work that runs on infrastructure the customer controls.
The pitch is simple: keep the AI next to the data. Forexploration and production (E&P) teams, that data includes seismic surveys,well logs, engineering reports and geological studies, much of it large andsome of it commercially sensitive.
A private AI layer on top of RiVA
AViR extends RiVA, GeoComputing’s existing cloud platform for petro-technical workflows. RiVA puts high-performance computing close to the data that geoscience applications depend on. AViR adds locally hosted language models to that environment, designed for private operation inside a customer’s own country.
That in-country design is aimed at data sovereignty, which the company lists alongside privacy as a core selling point. The company says AViR keeps teams in control of their environment while they use AI to search and interpret their own material.
What AViR is meant to do
GeoComputing lists three advantages. First, better knowledge retrieval: the system pulls insights from seismic data, well logs, engineering reports, geological studies and other technical documents, while keeping existing user access controls and data governance in place. That detail matters, because an AI assistant that ignores who is allowed to see which files would create a new risk rather than reduce one.
Second, proximity to workloads. By running models locally, alongside primary data and high-performance computing resources, AViR aims to limit data movement and lower latency. The company says this gives geoscientists, engineers and technical staff dedicated AI resources when they need them.
Third, cost predictability. Customers run inference on infrastructure they own instead of paying metered fees to outside AI services. As more teams across exploration, production and technical operations adopt AI, usage-based bills can climb quickly. A fixed infrastructure cost is easier to budget, though it also means the customer carries the hardware.
The argument behind the product
GeoComputing CEO John Creevan framed the launch around a question of location. “Where intelligence runs matters just as much as what it can do,” he said, according to the company’s announcement. He argued that energy organizations must keep control of their most valuable data while using AI tools, and said AViR lets E&P teams gain intelligence, reduce complexity and predict costs without giving up control, governance or accessibility. Creevan also described the product as the next chapter of private AI on the RiVA platform.
His claim that E&P teams can do this “for the first time ever” is the company’s own and has not been independently tested. No customers, pricing, hardware requirements or availability dates were included in the announcement.
What RiVA already promises
AViR will lean on the platform beneath it. GeoComputing saysRiVA addresses long-standing problems in geoscience computing: weak systemperformance, complex environments, huge data volumes, dispersed teams, limitedtechnical support and high deployment costs. The company says itshigh-throughput architecture and streamlined deployment cut processing times,so tasks that once took days can finish in hours. RiVA also includes anintegrated disaster recovery option intended to protect business continuity anddata resilience. Those figures are also the company’s own.
GeoComputing was founded in 2004 and describes itself as an IT consultancy for oil, gas, energy and other geoscience companies. Its services cover solution design, project management, infrastructure, application support and data management, with round-the-clock support from subject-matter experts.
What to watch
The announcement leaves open how AViR will perform in practice. Buyers will want to know which models it runs, how well they handle dense technical documents, and what a customer must own to run it. Still, the direction is clear: in a sector where data is both an asset and a liability, vendors are betting that private, local AI will be an easier sell than a public service.
