skills.rest
terrene-foundation
terrene-foundationOFFICIAL
Explore all skills by this author →
WebsiteGitHub

kailash-ml

Orchestrate ML training, registry, monitoring, and ONNX serving via km.* APIs.

Updated Mar 31, 2026
Search Tags:#machine learning#reproducibility#model registry#automl#drift monitoring#onnx serving#polars-native
One-click install
npx skills add https://github.com/terrene-foundation/kailash-coc-claude-rb --skill kailash-ml-terrene-foundation
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kailash-ml
Source: https://github.com/terrene-foundation/kailash-coc-claude-rb/tree/main/.claude/skills/34-kailash-ml
Command: npx skills add https://github.com/terrene-foundation/kailash-coc-claude-rb --skill kailash-ml-terrene-foundation

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It prevents unstructured, error-prone ML workflows by enforcing a production-grade lifecycle for training, drift monitoring, registry management, and ONNX serving.

Core Features & Use Cases

  • Engine-first ML lifecycle API: Use the canonical async km.* verb surface (e.g., train, register, serve, diagnose, watch, resume) to keep workflows consistent.
  • Polars-native data correctness: Enforces polars-only engine interfaces to reduce conversion bugs and drift between development and production.
  • End-to-end institutional safety: Provides deterministic conversion boundaries (interop.py), an allowlist for model class strings, and agent guardrails for AutoML recommendations.
  • Model lifecycle + reproducibility: Supports staging → shadow → production → archived with SHA256 integrity verification and lineage graphs.
  • Monitoring + retraining triggers: DriftMonitor detects distribution shifts with KS/chi2/PSI/Jensen-Shannon tests and can schedule checks or trigger retraining.
  • Cross-language deployment: Exports verified ONNX models for serving (including metadata embedding) and exposes inference via Nexus-compatible patterns.

Quick Start

Ask the AI to train a model end-to-end using the km.* pipeline from polars data, then register it for serving and enable drift monitoring for future retraining decisions.

Frequently Asked Questions about kailash-ml

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I track machine learning model drift in production?▼

Machine learning model drift is tracked using statistical tests like KS, chi2, PSI, and Jensen-Shannon to detect distribution shifts, which can then automatically schedule checks or trigger model retraining.

What is a model registry lifecycle with reproducibility checks?▼

A model registry lifecycle with reproducibility checks manages models through staging, shadow, production, and archived stages, verifying integrity with SHA256 hashes and lineage graphs to ensure reproducible deployments.

How do I serve ONNX models with embedded metadata for inference?▼

ONNX models are served by exporting verified models with embedded metadata, exposing inference via Nexus-compatible patterns to enable cross-language deployment and reliable production serving.

Can I use AutoML with human approval gates and guardrails?▼

AutoML can be used with agent guardrails that enforce human approval gates, ensuring institutional safety by requiring manual validation before accepting automated model recommendations.

Does machine learning reproducibility require polars-native data inputs?▼

Machine learning reproducibility requires polars-native data inputs to enforce engine-only interfaces, reducing conversion bugs and drift between development and production environments.

What is the best way to orchestrate ML workflows from training to serving?▼

Orchestrating ML workflows from training to serving is best handled through an engine-first async API surface covering training, registration, serving, diagnosis, and monitoring to maintain consistent production pipelines.

Related Skills

Explore More Data & Analytics Skills→
OwnLabAIOwnLabAI
21

stable-baselines3

Train reinforcement learning agents with Stable Baselines3 algorithms.

Official
Advanced
zenobi-uszenobi-us
68

machine-learning-engineer

Deploy scalable ML models with real-time inference, monitoring, and CI/CD pipelines.

Official
Advanced
RHEcosystemAppEngRHEcosystemAppEng
48

model-monitor

Monitor deployed InferenceServices for bias and data drift using SPD, DIR, and statistical metrics.

Official
Intermediate
aRustyDevaRustyDev
1

ml-pipeline-workflow

Design and manage end-to-end MLOps pipelines with DAG-based orchestration.

Community
Advanced