Incident remediation company Resolve AI Inc. announced the launch of SRE-1, a new model trained to find root causes in running environments, SiliconANGLE reported.
Software development has seen sweeping automation from artificial intelligence, but running production systems remains largely untouched by these tools, according to company statements. Production environments lack the public training corpora that models use to learn published code, leaving incidents private to each firm.
SRE-1 is a 27 billion-parameter model built on open weights. A team of 16 AI researchers trained the model over a final run of 7.4 days using up to 512 graphics processing units, SiliconANGLE reported.
Reinforcement learning alone drove a rise in the pass rate for a single investigation from 6.9% to 38.8% when measured on identical incidents before and after training. Resolve AI built the model using Snowflake Inc.’s Cortex Training serverless service.

Primary on-Call Service
Primary On-call puts artificial intelligence agents on the critical path to handle pages ahead of human engineers. Each alert is investigated across telemetry, infrastructure, code, and operational knowledge.
Depending on investigation findings, the agent closes the alert, mitigates the issue where team policies permit, or escalates to a human engineer. The system then checks the fix against the production baseline until recovery holds.
Resolve AI offers a shadow mode for teams that are hesitant to hand over pagers immediately. This mode records the call the agent would have made on each incoming page without taking automated action.
Primary On-call handles up to 80% of paged alerts in production without requiring human escalation. Salesforce Inc. tested the system during a dark launch, where the agent closed more than 70% of the CRM Analytics team paging alerts without waking an engineer, according to Bhupesh Patel, software engineering principal architect at Salesforce.
SRE-1 Ensemble Matches Top Model at Lower Cost
Testing on 49 difficult live production cases, run five times each, showed an ensemble of five SRE-1 investigations matching the strongest frontier model tested at a lower cost, SiliconANGLE reported. The system solved about twice as many incidents as general-purpose coding agents across 178 production alert investigations.
“No engineering team can keep up with how fast code and patches land in production,” said Spiros Xanthos, chief executive officer at Resolve AI, telling SiliconANGLE that the answer requires agents accurate and fast enough to sit on the critical path.
Vivek Raghunathan, senior vice president of engineering at Snowflake, said Resolve AI acted as the first design partner for Cortex Training. Primary On-call is now generally available, while access to SRE-1 remains restricted to Resolve AI’s own platform.
Frequently Asked Questions
What funding backs Resolve AI?
Resolve AI is a venture capital-backed startup that has raised more than $190 million to date. This total includes a $40 million extension to its Series A round in April, which valued the company at $1.5 billion and was led by DST Global and Salesforce Ventures.
Which companies use Resolve AI?
Notable customers using the platform include Coinbase Global Inc., DoorDash Inc., Figma Inc., MSCI Inc., Nasdaq Inc., SAP SE, and Zscaler Inc., alongside early users like Salesforce Inc.
Resolve AI Inc. launched in September 2024 with the goal of automating production engineering through artificial intelligence. Founded by a team with a background in building observability tools and creating open-source projects like OpenTelemetry, the company emerged from the founders’ experiences with engineering burnout caused by intensive on-call duties. The organization operates ResolveAI Labs, an internal research division focused on developing agentic systems capable of troubleshooting and resolving complex production issues.
Where is SRE-1 accessible?
Access to the SRE-1 model is limited strictly to Resolve AI’s own platform rather than being available for external deployment.
Keep reading