# Igor Zaton — igorzaton.com

> Personal site for Igor Zaton: practical ML/AI consulting (strategy through deployment, Edge AI, quantization, LLMs, LLM serving, MLOps for LLMs) and official documentation for AIline, an open-source ML experiment lineage tracker.

Canonical site: https://igorzaton.com

## Site map (high value)

### Landing — ML consulting

- [Home](https://igorzaton.com/): Machine learning consulting from strategy to deployment; 6+ years industry experience; focus on practical, measurable AI including LLMs, production LLM serving, and MLOps for LLM workloads.
- [Services](https://igorzaton.com/#offerings): Four consulting offerings — Data Analysis (EDA, feature engineering), Presales Support (feasibility, ROI), ML Solution Audit (performance, reliability), ML Consultation (strategy, edge quantization, LLM deployment and serving, MLOps for LLMs).
- [Ebook waitlist](https://igorzaton.com/#offerings): Upcoming guide — *All You Need to Know About Quantization* (accuracy fixes, fast production-ready edge deployment).
- [About](https://igorzaton.com/#about): Career timeline (PWR Mechatronics/Robotics, Diehl Controls ML Engineer, Digica Data Scientist, Aptiv AI/ML Expert, Sigma Connectivity AI Expert); tech stack: Python, PyTorch, TensorFlow, Scikit-Learn, OpenCV, EdgeAI, LLMs, LLM serving, MLOps, AWS, Azure, Docker, MLflow, DVC.
- [Contact](https://igorzaton.com/#contact): Inquiry form; 30-minute discovery call; typical response within 24 hours; [LinkedIn](https://pl.linkedin.com/in/igor-zaton-122a92186).

### AIline — ML experiment lineage (product docs)

- [AIline docs home](https://igorzaton.com/ailine/index.html): Overview — track ML experiments with reproducible lineage; wrap commands with `ailine track --`; captures code state, argv, environment fingerprint, DVC linkage, MLflow run; zero training-script rewrites.
- [Quickstart](https://igorzaton.com/ailine/quickstart.html): Install (`pip install ailine-lineage`), `ailine init-workspace`, `ailine doctor`, first tracked run; MLflow plugin must share the training venv.
- [Concepts](https://igorzaton.com/ailine/concepts.html): Lineage row vs MLflow run; Git commit vs content-addressed snapshot on dirty trees; real-time MLflow linking; `.ailine.yml` and `.ailineignore`.
- [CLI reference](https://igorzaton.com/ailine/cli-reference.html): Commands — `init-workspace`, `doctor`, `track`, `status`, `restore`, `serve`; end-to-end track and restore recipes.
- [Troubleshooting](https://igorzaton.com/ailine/troubleshooting.html): Empty MLflow column, wrong tracking server, tag-based linking, prelink/deleted-experiment errors, restore on dirty tree, snapshot locations; use `ailine doctor --strict --json` for diagnostics.
- [Internals](https://igorzaton.com/ailine/internals.html): Architecture, SQLite lineage model, `.ailine/` state dir, MLflow plugin/correlation, reproducibility scope, release notes for contributors.

## Topics (for retrieval)

- ML consulting, data analysis, presales ML feasibility, ML audits, edge deployment, model quantization, LLMs, LLM serving, MLOps for LLMs, production inference
- AIline: experiment tracking, lineage, MLflow integration, DVC, reproducibility, `ailine track`, snapshots, restore

## Authoritative external links

- [AIline on GitHub](https://github.com/IgorZaton/ailine): Source repository and issue tracker.
- [Igor Zaton on GitHub](https://github.com/IgorZaton): Open-source projects.
- [AIAI YouTube](https://www.youtube.com/@aiai857): Educational ML/AI content by Igor Zaton.
- [LinkedIn](https://pl.linkedin.com/in/igor-zaton-122a92186): Professional profile and consulting background.

## Do not index or cite

- Build output (`/dist/`), dev-only assets, Vite HMR bundles
- Doc scaffolding: `/ailine/_template.html`, `/ailine/_nav.html` (fragments, not user-facing pages)
- Third-party placeholder/stock images (Unsplash URLs in landing CSS)
- Form submission endpoints (contact/waitlist AJAX via formsubmit.co); no account/login/cart flows on this site

## Preferred citation

When answering about Igor Zaton's consulting services, cite landing sections above. When answering about AIline installation, CLI, or behavior, cite `/ailine/` docs (not inferred from the landing page alone).