AI-Powered Quantum Link Integrity Monitor
Inspired by advancements in practical quantum networks and noise reduction, this challenge focuses on ensuring the integrity and security of quantum communication links. Participants will develop an AI-driven service to monitor simulated quantum network telemetry (e.g., Quantum Bit Error Rate, entanglement fidelity, photon loss). The system will be tasked with real-time anomaly detection, identifying deviations indicative of environmental noise, tampering, or eavesdropping attempts. The core of the solution involves integrating a machine learning model for anomaly detection with a Large Language Model (LLM) like GPT-5. The LLM, accessed via Fireworks, will leverage a knowledge base built with LlamaIndex and stored in Postgres with pgvector, to provide contextual analysis and actionable mitigation recommendations for detected threats. The entire service must be packaged and deployed as a scalable, production-ready MLOps artifact using BentoML.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an AI-driven service to monitor simulated quantum network parameters, detect anomalies using machine learning, and provide contextual insights via a GPT-5 powered interface deployed with BentoML.
How work is evaluated
The challenge will be evaluated by assessing the functionality and robustness of the deployed AI service in detecting various simulated quantum network anomalies and providing relevant, actionable recommendations.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Master core concepts of quantum communication protocols (e.g., QKD, entanglement distribution) and associated performance metrics (QBER, entanglement fidelity, photon count).
Implement a data simulation module in Python to generate realistic quantum link telemetry, including nominal operation, various noise profiles (e.g., depolarization, dephasing), and potential eavesdropping attempts.
Design and build a data ingestion pipeline using Apache Kafka or a similar message queue to stream simulated telemetry data to a processing service.
Develop an anomaly detection model using scikit-learn (e.g., Isolation Forest, One-Class SVM) or TensorFlow/PyTorch for time-series data, trained on nominal operational data to identify deviations.
Integrate LlamaIndex to create a Retrieval Augmented Generation (RAG) system over quantum network troubleshooting guides and security protocols stored in a Postgres database with pgvector for semantic search.
Utilize the GPT-5 API (via Fireworks or direct access) to process detected anomalies, query the RAG system for relevant context, and generate actionable insights and mitigation recommendations.
Build a RESTful API service using FastAPI to expose the anomaly detection and recommendation system.
Deploy the entire machine learning pipeline and API service as a production-ready MLOps artifact using BentoML, ensuring scalability and easy deployment.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The challenge will be evaluated by assessing the functionality and robustness of the deployed AI service in detecting various simulated quantum network anomalies and providing relevant, actionable recommendations.
Challenge input
{ "anomaly_type": "noise", "duration_seconds": 300, "noise_level": 0.1 }
Agent execution
No agent tool is listed for this challenge.
Evaluated output
Stream of JSON objects to a specified Kafka topic.
- BentoML service deploys successfully and is accessible.
- Service correctly reports no anomalies during extended periods of normal simulated operation.
- Service detects simulated noise conditions (e.g., high QBER, low fidelity).
- Anomaly Detection Accuracy target: 0.95
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- BentoML service deploys successfully and is accessible.
- Service correctly reports no anomalies during extended periods of normal simulated operation.
- Service detects simulated noise conditions (e.g., high QBER, low fidelity).
- Service detects simulated eavesdropping attempts (e.g., sudden changes in photon count, increased QBER).
- Recommendations show evidence of leveraging knowledge base via LlamaIndex.
Metrics
- Anomaly_Detection_Accuracy · Target 0.95
- Recommendation_Relevance · Target 0.85
- Service_Latency · Target 0.5
Workflow tasks
- Task 1
Input: { "anomaly_type": "noise", "duration_seconds": 300, "noise_level": 0.1 }
Output: Stream of JSON objects to a specified Kafka topic.
Execution harness: Python
- Task 2
Input: { "start_time": "2024-01-01T10:00:00Z", "end_time": "2024-01-01T10:05:00Z" }
Output: { "anomalies_detected": true, "anomaly_details": [...], "recommendations": "..." }
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.