Build a Thermal Throttling Incident Timeline and Operator Rebalancing Engine
Analyze time-series DCGM thermal logs during server throttling events. Hosted Claude reconstructs event sequences and drafts operator chassis airflow mitigation recommendations.
What you are building
The core problem, expected build, and operating context for this challenge.
Analyze time-series DCGM thermal logs during server throttling events. Hosted Claude reconstructs event sequences and drafts operator chassis airflow mitigation recommendations.
Datacenter Operations Engineer
Parse normalized DCGM time-series telemetry records including GPU temperatures, power draw, and throttle flags.
Reconstruct the chronological timeline of temperature breaches and clock frequency reductions across affected chassis slots.
Send chronological telemetry logs to hosted Claude to isolate thermal runaways and formulate remediation steps.
Produce a structured operator inspection runbook containing fan curves and workload throttling mitigations.
Web dashboard showing thermal excursion graphs and human operator sign-off panels.
Construct synthetic DCGM CSV files recording timestamps, device index, temperature, power draw, and throttling duration. Starter examples below are synthetic. They are not downloaded records.
NVIDIA DCGM exporter Consumes standardized DCGM exporter metric types such as GPU temperature and throttling reason indicators.
mlflow, anthropic, pandas, pydantic. MLflow tracks thermal event runs, metric parameters, and operational remediation artifacts across historical incidents. Hosted Claude API for narrative incident root-cause reconstruction and mitigation generation.
Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.
Correctly identifies the exact primary throttle timestamp and initial trigger temperature.
Flags secondary cascading device throttling events occurring within 120 seconds of the primary trigger.
Produces actionable mitigation runbook with rack slot locations verified by human sign-off.
```json [ { "input": { "chassis_id": "rack-4-blade-02", "telemetry_stream": [ { "timestamp": "2025-06-01T10:00:00Z", "gpu_idx": 0, "temp_c": 74, "throttle_active": false, "clock_mhz": 1800 }, { "timestamp": "2025-06-01T10:01:00Z", "gpu_idx": 0, "temp_c": 86, "throttle_active": true, "clock_mhz": 1150 }, { "timestamp": "2025-06-01T10:01:30Z", "gpu_idx": 1, "temp_c": 84, "throttle_active": true, "clock_mhz": 1200 } ] }, "expected_output": { "incident_detected": true, "primary_trigger": { "timestamp": "2025-06-01T10:01:00Z", "gpu_idx": 0, "temp_c": 86 }, "cascading_count": 1, "recommended_action": "INSPECT_BLADE_FAN_BAFFLE" } }, { "input": { "chassis_id": "rack-4-blade-05", "telemetry_stream": [ { "timestamp": "2025-06-01T11:00:00Z", "gpu_idx": 0, "temp_c": 68, "throttle_active": false, "clock_mhz": 1800 }, { "timestamp": "2025-06-01T11:01:00Z", "gpu_idx": 0, "temp_c": 70, "throttle_active": false, "clock_mhz": 1800 } ] }, "expected_output": { "incident_detected": false, "primary_trigger": null, "cascading_count": 0, "recommended_action": "MAINTAIN_CURRENT_COOLING" } } ] ```
Read-only simulated diagnostics. Does not adjust physical fan controllers or alter operational firmware settings. Use mock external writes. Require human review before any real deployment.
Participation guide
Read-only simulated diagnostics. Does not adjust physical fan controllers or alter operational firmware settings. Use mock external writes and human review.
How work is evaluated
Correctly identifies the exact primary throttle timestamp and initial trigger temperature. Flags secondary cascading device throttling events occurring within 120 seconds of the primary trigger. Produces actionable mitigation runbook with rack slot locations verified by human sign-off.
Runnable source, pinned dependencies, setup commands, working interface, example outputs, recorded demo, tests, and action logs.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
Complete workflow
Parse normalized DCGM time-series telemetry records including GPU temperatures, power draw, and throttle flags. Reconstruct the chronological timeline of temperature breaches and clock frequency reductions across affected chassis slots. Send chronological telemetry logs to hosted Claude to isolate thermal runaways and formulate remediation steps. Produce a structured operator inspection runbook containing fan curves and workload throttling mitigations.
Missing or incorrect.
Partly complete, with documented gaps.
Complete, with reproducible evidence.
Correctness and evidence
Correctly identifies the exact primary throttle timestamp and initial trigger temperature. Flags secondary cascading device throttling events occurring within 120 seconds of the primary trigger. Produces actionable mitigation runbook with rack slot locations verified by human sign-off.
Missing or incorrect.
Partly complete, with documented gaps.
Complete, with reproducible evidence.
Review interface
Web dashboard showing thermal excursion graphs and human operator sign-off panels.
Missing or incorrect.
Partly complete, with documented gaps.
Complete, with reproducible evidence.
Safe execution
Read-only simulated diagnostics. Does not adjust physical fan controllers or alter operational firmware settings.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Parse normalized DCGM time-series telemetry records including GPU temperatures, power draw, and throttle flags.
Reconstruct the chronological timeline of temperature breaches and clock frequency reductions across affected chassis slots.
Send chronological telemetry logs to hosted Claude to isolate thermal runaways and formulate remediation steps.
Produce a structured operator inspection runbook containing fan curves and workload throttling mitigations.
Reference links and supporting material
Construct synthetic DCGM CSV files recording timestamps, device index, temperature, power draw, and throttling duration. Supplied starter examples are synthetic, not source extracts.
Challenge setup
Configuration summary for this challenge.
Challenge input
Input details are not configured.
MLflow
MLOps
Evaluated output
No output checks are configured.
- Parse normalized DCGM time-series telemetry records including GPU temperatures, power draw, and throttle fl...
- Correctly identifies the exact primary throttle timestamp and initial trigger temperature. Flags secondary...
- Web dashboard showing thermal excursion graphs and human operator sign-off panels.
- 2 public reference cases
- 1 configured tool
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
- MLflow · Optional
Evaluation contract
- Complete workflow · Weight 40
- Correctness and evidence · Weight 30
- Review interface · Weight 20
- Safe execution · Weight 10
Reference cases
- 2 public reference cases
Workflow tasks
No workflow tasks are configured.
Recipe freeze marker
No recipe freeze time is recorded.