Building your own DIY agent for incident resolution?

→

Understand what to look out for.

Building your own DIY agent for incident resolution?

→

Understand what to look out for.

Building your own DIY agent for incident resolution?

→

Understand what to look out for.

Customer success story · Deutsche Bahn

Deutsche Bahn: faster incident diagnosis inside strict data-handling boundaries

Hyground operated inside Deutsche Bahn's environment to accelerate root-cause analysis in passenger information systems without sending operational data outside the VPC. 8-week proof of concept, DB Reisendeninformation, distributed Kubernetes environment.

Book a meeting

Download PDF

The challenge

Distributed Operational Friction

Deutsche Bahn’s passenger information systems operate in a distributed cloud environment with multiple Kubernetes clusters, large telemetry volumes, and strict requirements for reliability and data handling.

“In stressful on-call situations, Hyground helped us access the exact system knowledge we needed much faster.” PoC Team Member, DB RIS

Alert overload

Engineers had to work through large volumes of warnings and telemetry across multiple clusters before isolating the real source of failure.

Explore Incident Investigation

➞

Knowledge concentration

Not every on-call engineer had the same system depth, which increased escalation and investigation time during critical incidents.

Data-handling boundaries

Sensitive operational data could not be sent to external cloud AI providers for analysis.

Explore Security

➞

The solution

Perimeter-Bounded Autonomy

Hyground was deployed in a dedicated sandbox cluster inside Deutsche Bahn’s environment and connected to the existing toolchain with read-only investigation paths.

Customer-controlled deployment

Deployment via Helm chart and integrated with systems such as OpenSearch, Prometheus, and Kubernetes without changing the production operating model.

Context-aware investigation

Hyground used telemetry and internal context to accelerate root-cause analysis and surface the relevant system knowledge for the on-call engineer.

Human-governed workflow

Engineers received evidence-based findings and retained control over any production decision or action.

Results

Sovereign SRE Enablement

“Based on the positive results, the rollout expanded toward additional business-critical applications and clusters.” PoC Lead, DB RIS

Under 5 min

Faster investigation

The proof of concept materially reduced investigation time, including workflows in which root-cause assessment was produced in under five minutes.

Up to 85%

Up to 85% lower MTTR

Across the tested workflows, mean time to resolution dropped by as much as 85%, with deployment confined inside Deutsche Bahn's VPC.

Broader team enablement

System knowledge became more accessible across the SRE team, reducing dependence on a small number of subject-matter experts.

Control preserved

Operational data remained inside Deutsche Bahn's VPC, supporting the organisation's data-handling requirements.

Evaluate the same operating model in your environment

Book a meeting

Reach out