I Built an AI That Designs MLOps Infrastructure (Without Letting AI Generate YAML)

Short summary

DeployCraft AI is a tool that generates deployment blueprints from high-level inputs while keeping infrastructure generation deterministic through a rules engine and Jinja2 templates. The LLM is restricted to explaining architecture decisions rather than generating YAML, improving reliability. The project includes architecture scoring, graceful fallbacks for LLM failures, and packages all artifacts into a downloadable ZIP.

• •Three-layer architecture: rules engine for deterministic blueprint, templates for YAML, LLM only for explanations
• •AI explains architecture choices (e.g., why Cloud Run) but never generates deployment manifests directly
• •Includes deterministic architecture scoring, graceful LLM fallbacks, and ZIP packaging of all artifacts

Generated with AI, which can make mistakes.