MLOps Governance Dashboard
PUBLIC CODE VERIFIEDMonitoring, drift (PSI/KS), data quality, audit events, and governance-oriented MLOps workflows.
- monitoring
- data drift
- PSI
- KS
- data quality
- audit
- alerting
- observability
- MLOps
Monitoring, drift (PSI/KS), data quality, audit events, and governance-oriented MLOps workflows.
Project Autopsy
Same project, six engineering perspectives. Switching lenses keeps RUN and X-RAY mounted so lab/scrubber state is not wiped.
Monitoring, drift (PSI/KS), data quality, audit events, and governance-oriented MLOps workflows.
Deterministic Runtime Lab — PORTFOLIO_EXTENSION. Prefer this lens over a separate route when you want lab state to survive inspection switches.
Monitor · healthy · phase idle · action baseline · seed 42 · n=200
PSI and KS severities currently agree (or one metric is unavailable).
Accessible bin counts for reference vs current — not a Grafana panel.
| Bin | Range | Reference count | Current count | Ref % | Cur % |
|---|---|---|---|---|---|
| 0 | [-3.28, -2.71) | 1 | 1 | 0.5 | 0.5 |
| 1 | [-2.71, -2.14) | 2 | 2 | 1.0 | 1.0 |
| 2 | [-2.14, -1.57) | 4 | 4 | 2.0 | 2.0 |
| 3 | [-1.57, -1.00) | 26 | 26 | 13.0 | 13.0 |
| 4 | [-1.00, -0.42) | 37 | 37 | 18.5 | 18.5 |
| 5 | [-0.42, 0.15) | 40 | 40 | 20.0 | 20.0 |
| 6 | [0.15, 0.72) | 36 | 36 | 18.0 | 18.0 |
| 7 | [0.72, 1.29) | 33 | 33 | 16.5 | 16.5 |
| 8 | [1.29, 1.86) | 12 | 12 | 6.0 | 6.0 |
| 9 | [1.86, 2.43) | 9 | 9 | 4.5 | 4.5 |
Project X-Ray — responsibility layers with isolation, component selection, and Source Trace. Causal reconstruction (M9) remains below.
Project X-Ray
Responsibility-layer inspection grounded in public evidence — not invented Redis/MLflow or fake production service maps.
What is portfolio simulation versus public code. Keeps the Runtime Lab honest.
Monitoring concepts as represented in the public repo — no fabricated Grafana dashboards.
Drift detection utilities: PSI and KS paths in public source files.
Audit recording and policy checks as separate responsibilities.
Causal reconstruction (M9)
Architecture reasoning reconstruction
Causal reconstruction of how the MLOps Governance system hangs together — not a fake historical v0 → v1 → v2 evolution.
The public MLOps Governance Dashboard is code-verified. The browser Runtime Lab is a PORTFOLIO_EXTENSION grounded in those concepts — not the exact historical production runtime.
Why this stage · Start with the honesty boundary so later stages cannot be mistaken for live production infrastructure.
Revealed stages: System Boundary.
Engineering decisions grounded in public evidence — not a fabricated post-mortem diary.
Public drift utilities expose both Population Stability Index and Kolmogorov–Smirnov paths so severity disagreement can be inspected rather than hidden behind one number.
Audit event representation and policy/governance checks are distinct modules — recording what happened is not the same as gating a decision.
The inspectable lab recomputes real local math grounded in public sources, but it is not claimed as the exact historical production runtime.
Exhibits appear here only when they clear the evidence gate. Future renderer is ready; the canonical dataset is currently empty.
Deterministic portfolio simulation grounded in verified MLOps concepts — not the original production runtime.
PORTFOLIO_EXTENSION concept node — browser lab grounded in verified MLOps concepts; not claimed as original production runtime.