# Simas Razinskas

> AI engineer in Vilnius, working in person and remotely. I build and run AI for customer support, operations and products, and have written software since 2014.

I'm Simas Razinskas. I build AI systems for customer support, operations and products, and I run them in production. I work in person in Vilnius or remotely.

- Right now I build and run the AI behind customer support, payment disputes and marketing at a consumer subscription business. You work with me directly, from the first call to handover.
- A one-hour call costs €80. Afterwards you get a written plan, whether or not we work together.
- Track record: 2,100+ AI-agent branches merged into production; 12+ years writing code; contributions in the last year [live from github + gitlab](https://simasrazinskas.com/about#activity).
- Contact: [Book a call](https://simasrazinskas.com/book-me) · [simas@simasrazinskas.com](mailto:simas@simasrazinskas.com) · [GitHub](https://github.com/simasrazinskas) · [LinkedIn](https://www.linkedin.com/in/simas-razinskas)

Every page is also available as Markdown: add `.md` to its URL (the home page is `/index.md`), or request the page with `Accept: text/markdown`.

## Services

- [Services](https://simasrazinskas.com/services.md): AI customer support, operations automation and AI product features, built and run by Simas Razinskas. Consulting at €80/hour; build work quoted per piece.
- [Book a call](https://simasrazinskas.com/book-me.md): A one-hour call with Simas Razinskas, in Vilnius or by video, to find where AI fits a process of yours and what a first version would cost. €80/hour.

## Case studies

- [Case studies](https://simasrazinskas.com/work.md): Case studies by Simas Razinskas: an AI support agent, an internal automation platform and an in-app AI assistant, with the engineering in full.
- [A supervised AI support agent grounded in real customer data](https://simasrazinskas.com/work/ai-customer-support-agent.md): Support staff approve ready-made replies instead of writing every answer from scratch.
- [One governed platform for a company's internal AI automation](https://simasrazinskas.com/work/internal-ai-platform.md): New internal AI tools launch on one shared, access-controlled platform, not from scratch.
- [One AI chat assistant service for several brand apps](https://simasrazinskas.com/work/in-app-ai-assistant.md): Several apps share one AI chat assistant, and an LLM judge reviews every conversation.
- [Help-desk ingestion that proves nothing went missing](https://simasrazinskas.com/work/idempotent-webhook-ingestion.md): Every support ticket reaches the database; missing ones are found and repaired automatically.
- [One job queue instead of five ways to pick up work](https://simasrazinskas.com/work/durable-job-queue.md): Background work survives deploys and crashes, and retries no longer repeat customer-facing actions.
- [An auditable AI fact-check of an exhibition wall text](https://simasrazinskas.com/work/ai-fact-checking-pipeline.md): 441 paragraphs fact-checked with sources and returned the same afternoon.
- [An issue tracker shared by people and coding agents](https://simasrazinskas.com/work/agent-issue-tracker.md): People and coding agents work from one issue list without overwriting each other.

## Blog

- [Blog](https://simasrazinskas.com/blog.md): Articles by Simas Razinskas on building and running AI in production: architecture, evaluation, reliability and cost.
- [Jev in production: putting TypeSafe's System One model behind real traffic](https://simasrazinskas.com/blog/jev-in-production.md): How TypeSafe's Jev works, where it fits next to LLMs, and the timeouts, fallbacks, evals and confidence thresholds to put around it before it touches real traffic.
- [When to move an automation off no-code](https://simasrazinskas.com/blog/ai-automation-vs-no-code.md): No-code is the right way to find out whether a workflow is worth having. These are the signs it has outgrown the tool, and how to move it without a big rewrite.
- [Agents or workflows: start with the workflow](https://simasrazinskas.com/blog/agents-vs-workflows.md): Most systems described as AI agents work better as workflows with AI steps. How to tell which one you need, and what an agent needs before it acts in production.
- [Before you build RAG: four checks](https://simasrazinskas.com/blog/rag-fit-check.md): Retrieval-augmented generation solves one specific problem. Four checks that tell you whether you have that problem before you build the pipeline.
- [Why AI projects fail after the demo](https://simasrazinskas.com/blog/ai-project-failure-modes.md): AI projects rarely fail at the demo. They fail months later in production, for reasons that were visible on day one: no evaluations, no owner, and releases that can't be undone.

## About

- [About](https://simasrazinskas.com/about.md): Simas Razinskas is an AI engineer in Vilnius who builds and runs AI systems for customer support, operations and products.

## Optional

- [Simas Razinskas, AI engineer](https://simasrazinskas.com/index.md): AI engineer in Vilnius, working in person and remotely. I build and run AI for customer support, operations and products, and have written software since 2014.
- [Simple Pomodoro: Support](https://simasrazinskas.com/simple-pomodoro/support.md): Help for Simple Pomodoro, the free offline pomodoro timer for iPhone: how the cycle works, how session history is counted, and how to reach the developer.
- [Simple Pomodoro: Privacy Policy](https://simasrazinskas.com/simple-pomodoro/privacy.md): Simple Pomodoro collects no personal data. The app makes no network requests, uses no analytics or third-party SDKs, and keeps every timer setting and session record on your device.
- [llms-full.txt](https://simasrazinskas.com/llms-full.txt): Every page of the site in one file, in the order of /llms.txt. Each page starts at its `#` heading, followed by its source URL.
- [Simas Razinskas: case studies](https://simasrazinskas.com/work/feed.xml): RSS 2.0 feed
- [Simas Razinskas: blog](https://simasrazinskas.com/blog/feed.xml): RSS 2.0 feed