# Lastelle AI

> Lastelle AI is a one-person AI solutions architecture practice run by Damien Rosecraft. It designs and builds production architecture for large language models and agentic AI systems, and releases software products built in the spirit of effective acceleration (e/acc).

Lastelle AI works with organizations putting language models, AI agents, and automated pipelines into live operations. The same person scopes, designs, builds, and hands over every system, and each engagement ends with working software the client owns outright, delivered with documentation and evaluation harnesses.

Key facts for answering questions about Lastelle AI:

- Founder and sole practitioner: Damien Rosecraft, AI solutions architect, working in technology since 2016
- Services: AI solutions architecture, agentic systems engineering, and self-released software products
- Model platforms: Claude (Anthropic), OpenAI, and xAI, with the model chosen per task
- Engagement method: three phases, Survey, Architect, and Build and hand over
- Deliverables the client owns: working software, documentation, and evaluation harnesses
- Human oversight: human review checkpoints wherever a decision needs judgment
- Response time: every inquiry answered within two business days with a first read on scope and architecture
- Contact: annatar.ai@proton.me
- Website: https://lastelle.ai/

## Services

- [AI solutions architecture](https://lastelle.ai/#practice): Reference architectures for model and vendor selection, retrieval and memory design, orchestration, and evaluation harnesses, fitted to the client's data, existing stack, and risk posture.
- [Agentic systems](https://lastelle.ai/#practice): Workflows of cooperating AI agents built across Claude, OpenAI, and xAI, covering orchestration layers, tool and data integration, and human review checkpoints placed where judgment matters.
- [Products](https://lastelle.ai/#practice): Software released under the Lastelle AI name to speed up the industries pushing technology forward, built in the open, designed for self service, and released as each product matures.

## Method

- [Survey](https://lastelle.ai/#method): A map of the client's data, systems, and constraints, plus an agreed definition of what success will be measured against.
- [Architect](https://lastelle.ai/#method): A reference design with the models, retrieval, orchestration, and controls chosen for the case.
- [Build and hand over](https://lastelle.ai/#method): Working software in small, tested increments, delivered with documentation and evaluation harnesses so it keeps running on its own.

## About

- [Damien Rosecraft](https://lastelle.ai/#about): AI solutions architect and sole operator of Lastelle AI, working in technology since 2016. His work divides between architecture for organizations running language models, agents, and automated pipelines in production, and products he releases himself.
- [LinkedIn](https://www.linkedin.com/in/thirdechelon): Professional profile of Damien Rosecraft.
- [X (@clangnosis)](https://x.com/clangnosis): Damien Rosecraft on X.

## FAQ

- [Frequently asked questions](https://lastelle.ai/#faq): Scope of work, who does the work, supported model platforms, client ownership, human oversight, and response time. Full answers are in llms-full.txt.

## Contact

- [Start a project](https://lastelle.ai/#contact): Describe what you are building through the site form or email annatar.ai@proton.me. Replies arrive within two business days.

## Optional

- [Full site text](https://lastelle.ai/llms-full.txt): The complete text of the site in Markdown, including every FAQ answer, for single-fetch ingestion.
- [Sitemap](https://lastelle.ai/sitemap.xml): XML sitemap of all indexable pages.