SEOJEV / Laya
Autonomous technical SEO audit agent using graph signals and local Apple Silicon MLX inference
The Problem
SEO audits produce massive spreadsheets of disconnected crawl issues that sit unprioritized and unexecuted by development teams.
Engineering Approach
Combined site graph topology analysis with local on-device LLM reasoning to translate technical audit findings into prioritized engineering work orders.
What I Built & My Contribution
Engineered crawling engine, NetworkX site link graph builder, local MLX inference wrapper, and automated work-order generator.
Architecture & Data Pipeline
Crawl Engine → Link Graph Construction → Evidence Aggregation → Local MLX Model → Verifiable Work Orders
[ Target Website URL ]
│
▼
┌───────────────────────────────┐
│ Crawl & Graph Engine │
│ - NetworkX PageRank/Depth │
│ - Broken Links & Redirects │
└───────────────┬───────────────┘
│
▼
┌───────────────────────────────┐
│ Local MLX Inference │
│ - Structured Issue Grouping │
│ - Engineering Work Orders │
└───────────────────────────────┘Verified Outcomes & Metrics
Local private inference with zero cloud token cost; automated page depth and link graph calculations; actionable developer work orders.
Verified Implementation
The FastAPI application wires run workers, schedules, site artifacts, and dedicated work-order API routes. Verification here represents code audit and local inspection, distinguishing functional code paths from external cloud provider accounts.
Boundaries & Known Limitations
Requires Apple Silicon hardware and local model weights. Repository claims do not promise ranking or organic search traffic guarantees.