ml-engineering
Master the principles for constructing reliable machine learning systems from data collection through deployment. Learn pipeline design patterns, feature engineering strategies, model validation checklists, and serving approaches—plus monitoring techniques to detect drift and trigger retraining.
ML Engineering teaches you to design reproducible ML pipelines with experiment tracking, feature stores, and model validation.
AI-generated summary based on this skill's SKILL.md
Install
irahardianto/awesome-agv/ml-engineering · repository language: JavaScript
git clone https://github.com/irahardianto/awesome-agv
cp -r awesome-agv/.agents/skills/ml-engineering ~/.claude/skills/ml-engineeringnpx skillfed install irahardianto/awesome-agv/ml-engineeringFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How to build machine learning pipelines with reproducibility?
ml-engineering teaches you to design end-to-end ML pipelines that are reproducible and maintainable. Key practices include version controlling data and code, containerizing environments, documenting hyperparameters, and using orchestration tools like Airflow or Kubeflow. Reproducibility ensures that experiments can be re-run with identical results, enabling reliable governance and collaboration across teams.
What are ML model training and deployment best practices?
ml-engineering covers the full lifecycle: establish experiment tracking with tools like MLflow or Weights & Biases to log metrics and artifacts, implement a model registry for versioning and governance, validate models against comprehensive checklists before production, and choose appropriate serving patterns—batch inference for offline workloads or real-time APIs for low-latency needs. Monitor performance continuously post-deployment.
How does ml-engineering address preventing training-serving skew?
ml-engineering emphasizes feature engineering strategies and feature store design patterns to eliminate training-serving skew. Centralize feature computation, document feature definitions rigorously, and validate that training and serving environments use identical feature logic. Monitor data drift and feature distributions in production to detect divergence early and trigger automated retraining when thresholds are exceeded.
What model validation and monitoring techniques does ml-engineering cover?
ml-engineering provides model validation checklists covering fairness evaluation across protected groups, data leakage prevention, and performance benchmarking. For production monitoring, learn to detect data drift, track model performance metrics over time, and set up automated retraining triggers. These practices ensure models remain reliable and fair as real-world data evolves.
How should ml-engineering guide feature documentation standards?
ml-engineering stresses comprehensive feature documentation as foundational to MLOps infrastructure. Document each feature's definition, data source, transformation logic, and expected distributions. This prevents confusion during model handoffs, enables reproducibility, and supports data governance. Well-documented features make it easier to detect drift, debug model issues, and maintain consistency across training and serving.
What serving patterns does ml-engineering recommend for inference?
ml-engineering helps you choose between batch inference for offline scoring of large datasets and real-time API serving for low-latency predictions. Consider edge deployment for latency-critical or privacy-sensitive workloads. Each pattern has trade-offs in throughput, latency, and infrastructure complexity; select based on your use case requirements and resource constraints.
SKILL.md
rendered from the published skill — quoted content, verbatim
ML Engineering Principles
Guidelines for building reliable, reproducible machine learning systems.
When to Invoke
- Designing ML pipelines (training, serving)
- Feature engineering and data preparation
- Model evaluation and validation
- MLOps infrastructure decisions
ML Pipeline Design
Stages
Data Collection → Feature Engineering → Training → Evaluation → Deployment → Monitoring
Principles
- Reproducibility — versioned data, code, and config. Same inputs = same model.
- Experiment tracking — every run logged (MLflow, W&B, Neptune).
- Feature stores — centralized feature computation, reusable across models.
- Model registry — versioned models with metadata, promotion workflow.
Feature Engineering
- Compute features once, reuse everywhere — feature store pattern.
- Training-serving skew prevention — same transformation code in training and inference.
- Feature documentation — every feature has description, source, freshness requirement.
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 1 file
.agents/skills/ml-engineering/SKILL.md