skills.rest
Dazlarus
DazlarusCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

ml-ops

Implements model registries, serving APIs, drift detection, and automated retraining pipelines for ML operations.

Updated Sep 2, 2026
Search Tags:#drift detection#mlops#model deployment#mlflow#model registry#model monitoring#retraining
One-click install
npx skills add https://github.com/Dazlarus/karl-code --skill ml-ops-dazlarus
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ml-ops
Source: https://github.com/Dazlarus/karl-code/tree/main/.agents/skills/ml-ops
Command: npx skills add https://github.com/Dazlarus/karl-code --skill ml-ops-dazlarus

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Managing machine learning models in production requires versioning, monitoring, drift detection, and retraining infrastructure that teams often build ad hoc. This Skill provides concrete implementation patterns for the full ML lifecycle from deployment to automated retraining. ## Core Features & Use Cases - Model Registry & Serving: Register versioned models with metadata and serve them through a FastAPI prediction endpoint with health checks. - Monitoring & Drift Detection: Track latency, error rates, and accuracy in production, and detect data drift using Kolmogorov-Smirnov tests and Population Stability Index. - Experiment Tracking & Retraining: Log training runs with MLflow and trigger automated retraining when performance drops or drift is detected. - Use Case: A team deploying a churn prediction model can register each version, serve predictions via API, monitor accuracy against live labels, and automatically retrain when feature distributions drift. ## Quick Start Set up a model registry and FastAPI serving endpoint for my trained classifier with drift monitoring and automated retraining triggers.

Frequently Asked Questions about ml-ops

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

FAQPage Schema
How do I deploy a machine learning model to production with Python?▼

Register the model in a versioned registry with metadata, then serve it through a FastAPI endpoint that loads models from the registry and exposes a /predict route. Include health checks and model listing endpoints for operational visibility.

How to detect data drift in production ML models?▼

Compare current feature distributions against reference training data using the Kolmogorov-Smirnov test and Population Stability Index. A KS p-value below 0.05 or PSI above 0.2 indicates significant drift requiring investigation or retraining.

What is a model registry and how do I build one?▼

A model registry is a centralized store that versions trained models alongside metadata like metrics, hyperparameters, and training data hashes. This Skill implements one using joblib for artifact storage and a JSON index for version lookup.

Does MLflow support tracking PyTorch model experiments?▼

Yes, MLflow logs parameters, metrics, and PyTorch models through mlflow.pytorch.log_model within a run context. You can then query runs by metric to identify the best-performing experiment.

When should I trigger automated model retraining?▼

Trigger retraining when monitored accuracy falls below a defined threshold, when drift is detected in input features, or after a configured time interval. Combine these signals rather than relying on schedules alone.

What metrics should I monitor for a production model?▼

Track average, p95, and p99 prediction latency, error rate, and accuracy when ground-truth labels become available. Set thresholds for each metric and alert when they are exceeded.

Related Skills

Explore More Software Engineering Skills→
balaraj74balaraj74

Machine Learning

Train classification and regression models with PyTorch and evaluate using standard metrics.

Community
Advanced
MaelwalserMaelwalser

deployment-patterns

Orchestrate deployment workflows and rollback procedures for production applications.

Community
Advanced
AreteDriverAreteDriver
4

monitor

Instrument JSON logs, metrics, health checks, and alerting for production services.

Community
Basic
davila7davila7
30.5k

ml-engineer

Implements production ML systems with PyTorch, TensorFlow, model serving, and monitoring.

Community
Intermediate