Inspiration
Spacecraft telemetry is high-dimensional and difficult to interpret in real time. We wanted to build a system that not only detects anomalies, but also explains them in a way operators can act on.
What it does
Telemetry Copilot detects anomalies in multi-channel telemetry using LSTM-based prediction, provides early warnings before failures, and explains which channel is behaving abnormally and why.
How we built it
We integrated Telemanom for anomaly detection with a Python inference worker, a Next.js dashboard for visualization, and a Copilot layer powered by Claude to generate real-time explanations.
Challenges we ran into
We adapted Telemanom to modern Python and TensorFlow environments, handled sequence alignment issues in time-series prediction, and designed the system to work effectively with anonymized telemetry data.
Accomplishments that we're proud of
We built an end-to-end system that combines anomaly detection, real-time visualization, and AI-driven explanations into a cohesive monitoring experience, bridging research models with a practical application.
What we learned
We learned that anomaly detection alone is not sufficient — interpretability and actionable insights are essential for real-world monitoring systems.
What's next for Telemetry Copilot
We plan to extend the system with multi-channel modeling, anomaly classification, and deeper root-cause analysis to support more advanced monitoring scenarios.
Log in or sign up for Devpost to join the conversation.