triscravello/running-readiness-ml

How triscravello/running-readiness-ml works

This repository is an educational, end-to-end running-readiness prediction system, spanning synthetic data preparation and leakage controls through group-safe model selection, feature gating, neural-network transfer learning, and a secured CPU inference API. The main user path is a validated prediction request through the service into a lineage-verified runtime model. Separate development experiments and deployment-integrity checks support that path. Relationships follow the README and shown imports/calls; unsampled subsystem details are represented conservatively.

Main entry points: feature_experiment.py Feature gate · train_models.py Group-safe training · data_pipeline.py Data pipeline · neural_network_experiment.py Denoising pretraining

Components 19

Data contracts

Data pipeline src/data_pipeline.py
Schema drift gate src/schema_drift_gate.py
Feature gate src/feature_experiment.py

Classical modeling

Group-safe training src/train_models.py
Evaluation and lock src/evaluate_models.py
Threshold analysis src/threshold_analysis.py

Neural modeling

Denoising pretraining src/neural_network_experiment.py
Transfer fine-tuning src/finetune_experiment.py

Serving integrity

Lineage verification src/level7_lineage.py
Serving manifest src/level7_manifest.py
Deployment packaging src/level7_package.py
API validation src/level7_service.py
CPU model runtime src/level7_runtime.py
Model and handoff artifacts artifacts

Model experiments

Tree capacity study src/capacity_experiment.py
Leakage proxy study src/leakage_proxy_experiment.py
Overfit study src/overfit_experiment.py

Other components

Raw data source
Prediction client

Connections 20

Raw data source → Data pipeline: provides data README.md
Data pipeline → Group-safe training: prepares splits README.md
Data pipeline → Schema drift gate: checks drift src/schema_drift_gate.py
Group-safe training → Evaluation and lock: evaluates models README.md
Evaluation and lock → Feature gate: locks handoff README.md
Group-safe training → Threshold analysis: supplies candidates src/threshold_analysis.py
Group-safe training → Tree capacity study: supplies models src/capacity_experiment.py
Feature gate → Leakage proxy study: audits proxies src/leakage_proxy_experiment.py
Feature gate → Denoising pretraining: approves features README.md
Denoising pretraining → Transfer fine-tuning: transfers encoder src/finetune_experiment.py
Feature gate → Transfer fine-tuning: audits handoff src/finetune_experiment.py
Denoising pretraining → Overfit study: provides training methods src/overfit_experiment.py
Transfer fine-tuning → Model and handoff artifacts: produces model handoff README.md
Model and handoff artifacts → Lineage verification: hash-verified inputs README.md
Lineage verification → Serving manifest: verifies lineage src/level7_manifest.py
Lineage verification → Deployment packaging: gates packaging src/level7_package.py
Model and handoff artifacts → CPU model runtime: loads model files README.md
CPU model runtime → Serving manifest: verifies hashes src/level7_runtime.py
API validation → CPU model runtime: predicts with src/level7_service.py
Prediction client → API validation: submits request README.md

Drawn by GitDiagram on November 9, 2026 from the repository's file tree, README and source files. Read as Markdown ·