Anomx

From data streams to system intelligence

Anomx brings agentic AI, time-series analysis and anomaly detection into one connected workspace. It helps researchers and engineers discover unusual behavior, investigate what changed and turn complex operational data into findings they can act on.

Machines generate data. Anomx turns it into understanding. Connect signals to the systems that produce them, combine machine learning with an AI agent, and move from isolated measurements to a shared picture of your infrastructure.

An AI partner for your entire workflow

• Investigate in natural language. Explore data, inspect channels, develop analyses and work with models through an agent that understands the platform's tools and context.
• Find the unexpected. Combine forecasting, reconstruction and representation methods to reveal deviations in individual signals and across many dimensions.
• Keep intelligence running. Give background agents ongoing instructions to revisit connected data, investigate changes and carry useful context into the next run.
• Connect the whole system. Bring channels, datasets, analysis jobs, model artifacts and findings together with a navigable system hierarchy.
• Turn evidence into action. Review findings and recommendations, inspect execution history, and choose the boundaries within which agents can make platform changes.

Built for the complexity of research

From accelerator diagnostics and scientific instruments to industrial equipment and distributed infrastructure, Anomx is designed for environments where understanding a change matters as much as detecting it.

Its research context at DESY brings the challenges of scientific control systems directly into the product: heterogeneous signals, connected subsystems, continuous data acquisition and the need to trace an observation back to its source. Modular connectors and distributed workers connect analysis with the infrastructure and compute resources around it.

Research context: a schematic view of connected systems at DESY and European XFEL.

Three ways to recognize the unexpected

Forecasting learns what should happen next and highlights departures from that expectation. Reconstruction learns normal patterns and exposes observations that do not fit. Representation explores relationships in feature space to identify unusual combinations that individual thresholds can miss.

The reusable Python foundation provides datasets, models, scorers and detectors, with optional integrations such as Darts and PyTorch. Researchers can use these building blocks independently or bring them into the wider Anomx workflow.

Your infrastructure. Your intelligence.

Anomx is primarily designed for on-premises deployment. Bring the platform to your data, connect your own systems, and place acquisition and compute workers where they are needed. The agent supports a choice of model backends, including institutional services and local models.

A hosted SaaS version is coming soon, offering another way to access Anomx. Public platform onboarding and getting-started instructions are in preparation.

Start with the open-source agent

The Apache-2.0 licensed Anomx Python package and CLI agent are available independently of the proprietary Anomx platform. Use the agent for coding, data analysis and anomaly investigation from your terminal; connect it to a platform instance when you need shared context and orchestration.

pip install anomx
anomx

Explore Anomx · Meet the agent · Python package

For deployment discussions and early access, contact hello@anomx.io.