# Build Production ML Systems

Production machine learning requires reliable architecture, deployment, testing, and monitoring. This skill provides practical guidance across the complete ML system lifecycle.

## Install

```bash
npx skillstore add sickn33/ml-engineer
```

## Metadata

- Status: approved
- Slug: sickn33-ml-engineer
- Skillstore revision: r2
- Version status: missing
- Tree hash: c093e0d3a2a9dcdf539848cd745beb477496ec6e91903e7c2f48aad899506516
- Author: sickn33
- GitHub username: sickn33
- License: MIT
- Repository: https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/ml-engineer
- Ref: 81e05e636292629114b76cbb3922fbe57672fc02
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/sickn33-ml-engineer
- Manifest: https://skillstore.pages.dev/api/skills/sickn33-ml-engineer/manifest

## Capabilities

- Designs training, serving, feature, and monitoring architectures for production ML workloads.
- Recommends frameworks for distributed training, experiment tracking, orchestration, and model registries.
- Plans real-time, batch, cloud, edge, and container-based model deployment.
- Defines evaluation, A/B testing, drift detection, and model lifecycle strategies.
- Suggests performance and cost optimizations such as batching, caching, quantization, and autoscaling.
- Creates testing plans and operational runbooks for data, models, and ML services.

## Use Cases

- Plan a first production deployment: Select a serving pattern, deployment platform, monitoring plan, and rollback strategy for an existing model.
- Scale an ML platform: Design distributed training, feature serving, experiment tracking, and continuous delivery for growing workloads.
- Improve model operations: Define drift alerts, business metrics, retraining triggers, and controlled model rollout procedures.

## Prompt Templates

### Choose a deployment approach

```
Recommend a deployment approach for my [model type]. It serves [traffic] with [latency target] on [platform]. Explain tradeoffs and validation steps.
```

### Design a monitored inference service

```
Design a [batch or real-time] inference service for [workload]. Include APIs, scaling, failure handling, metrics, alerts, testing, and rollback.
```

### Create an experiment and rollout plan

```
Create an A/B test for [model change] using [business metric]. Define assignment, sample sizing, guardrails, significance checks, rollout, and stopping rules.
```

### Architect an end-to-end ML platform

```
Architect an ML platform for [use case] at [scale]. Cover data, features, training, registry, serving, observability, governance, cost, and disaster recovery.
```

## Limitations

- Guidance requires environment-specific validation before production use.
- The skill does not access infrastructure, train models, or deploy services by itself.
- Recommendations depend on complete requirements for data, scale, latency, cost, and compliance.
- Specialized legal, privacy, security, and statistical decisions still require qualified review.

## Best Practices

- Provide workload scale, latency, cost, data, platform, and compliance constraints.
- Validate recommendations with representative data, load tests, failure tests, and staged rollouts.
- Measure business outcomes with model quality, reliability, fairness, and infrastructure metrics.

## Anti Patterns

- Do not optimize model accuracy while ignoring latency, reliability, and business impact.
- Do not deploy without versioning, observability, rollback procedures, and ownership.
- Do not automate retraining without data validation, evaluation gates, and approval controls.

## Security Audit

- Audited at: 2026-08-04T14:27:57.689\+00:00
- Summary: Both static findings are false positives caused by ordinary Markdown prose. The skill contains no executable shell command, system reconnaissance behavior, or semantic evidence of malicious intent.

## Stats

- Views: 163
- Downloads: 12
- Favorites: 0
- Popularity score: 0