Skip to main content
Loading navigation
Back to challenges

Pipeline Telemetry Anomaly Detection Agent using Google ADK and Hugging Face

Crude oil pumped to export hubs must maintain strict pressure and flow metrics to prevent leaks or theft during the 3-month export window. Build a real-time pipeline monitoring agent using Google Agent Development Kit (ADK) with custom anomaly detection models hosted on Hugging Face Inference API. The agent processes stream telemetry (SCADA flow rates, pressure differentials, temperature spikes), invokes time-series models via Hugging Face API, and triggers emergency automated valve isolation when anomalies breach sovereign threshold standards.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Implement a pipeline SCADA anomaly detection and action agent using Google ADK and Hugging Face Inference API.

How work is evaluated

Evaluation

Evaluates pipeline anomaly detection speed and accuracy from telemetry logs.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

detect_sudden_leak

Ensure pressure drop correctly triggers leak alarm.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

anomaly_precision

Precision of leak detection against baseline noise. • target: 0.96 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Build pipeline telemetry processing agents using Google ADK.

  • Integrate Hugging Face Inference API for custom ML classification models.

  • Execute automated control tool calls upon anomaly confirmation.

Resources

Dataset notes

Time-series pipeline telemetry dataset covering normal operation and leak anomaly scenarios.

Challenge setup

Evaluates pipeline anomaly detection speed and accuracy from telemetry logs.

Challenge input

JSON containing pipeline_section, pressure_psi_stream, flow_bph_stream

Google ADK

Google's native agent SDK for building task-oriented systems.

Hugging Face Inference API

Direct API access to fine-tuned time-series ML models.

Evaluated output

JSON with anomaly_detected, anomaly_type, isolation_recommended

Evaluation checks
  • Ensure pressure drop correctly triggers leak alarm.
  • Precision of leak detection against baseline noise. • target: 0.96 • range: 0-1
Configured evaluation evidence
  • Benchmark: IoT-Anomaly-Bench
  • Anomaly Precision target: 0.96
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Google · Required
  • Hugging Face · Optional
  • Google ADK · Optional

Evaluation contract

  • detect_sudden_leak · Weight 1
  • anomaly_precision · Weight 1

Metrics

  • anomaly_precision · Target 0.96

Benchmarks

  • IoT-Anomaly-Bench
  • energy

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing pipeline_section, pressure_psi_stream, flow_bph_stream

    Output: JSON with anomaly_detected, anomaly_type, isolation_recommended

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Architecture
  • Hugging Face Inference API · Model Runtime

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Pipeline Telemetry Anomaly Detection Agent using Google ADK and Hugging Face