Gontse Sebolai/JHB/2026
Project laboratory

Synthetic model observability · Python proof

Lumen
Observatory

A bounded model-monitoring platform built with synthetic network-service telemetry, keeping lineage, drift and limitations visible.

Python 3.14FastAPI 0.139scikit-learn 1.9PostgreSQL 17Prometheus
Gontse portrayed head-to-toe as a Renaissance natural philosopher calibrating an armillary sphere and marking a celestial chart
Full-scene identity study · period clothing, working hands and environment painted as one composition.
09

Python tests

API, scoring, drift and asset contract

91.76%

statement coverage

Measured by the final local pytest run

0.9235

synthetic ROC AUC

Generated 1,100-sample holdout only

02

responsive captures

1,440px desktop and 390px mobile

Govern the measurement, not only the model.

The slice is deliberately small: one safe telemetry problem, one reproducible artefact and enough surrounding evidence to challenge it.

  1. 01Generate

    Seed 404 creates bounded infrastructure telemetry with controlled label noise.

  2. 02Train

    A random forest is accepted only when generated holdout ROC AUC reaches 0.85.

  3. 03Score

    FastAPI validates the reading and persists probability, band, signals and version.

  4. 04Observe

    Population stability and Prometheus metrics expose changes around the model.

Stack selected from employer evidence

The third proof changes the conversation.

After .NET and Java, Lumen responds to Python, data, model-monitoring and cloud-service signals in target banking and telecommunications roles without becoming another oversized flagship.

01

Python 3.14 + FastAPI 0.139

Typed APIs and a second backend ecosystem beyond .NET and Java.

02

scikit-learn 1.9

A reproducible training path, model artefact and explicit release gate.

03

PostgreSQL 17

Every synthetic reading keeps its inputs, output and model version together.

04

PSI + Prometheus

Input-distribution change and service behaviour remain inspectable.

05

Docker + GitHub Actions

A non-root runtime and pinned verification workflow.

Generated evidence

Training and seeded interface data are visibly synthetic.

Version lineage

Each score records the model version that produced it.

Drift visibility

PSI compares recent inputs with the training baseline.

Durable readings

PostgreSQL keeps the measured input and result together.

Verified locally; bounded before it is impressive.

Ruff, strict mypy and nine pytest checks passed locally on 07 September 2026 with 91.76% statement coverage under a targeted warning policy. Project-owned deprecations are promoted to errors; one understood third-party warning is filtered. Docker/PostgreSQL readiness, Prometheus output and one complete desktop/mobile scoring flow remain the separate 15 July 2026 checkpoint.

Generated infrastructure telemetry only; no person or customer data

Contextual signals orient investigation but are not causal explanations

Holdout metrics do not claim performance on a real telecommunications network

Local repository and noindex case study; no public or cloud deployment implied

Discuss the build View Aegis Ledger

Local repository and noindex; not presented as a deployed system.

Quality gates refreshed 07 September 2026 · Docker/browser checkpoint 15 July 2026.