# Tom Seidel - Java Freelancer & Cloud-Native AI Specialist

> Freelance Java consultant with 20+ years of experience in cloud-native architectures, microservices, DevOps, and AI-powered agentic systems. Based in Germany, remote-first.

Tom Seidel is a freelance consultant specializing in Java development, cloud-native architectures, AI integration, and DevOps. He supports projects end-to-end: from initial concept and architecture to production-ready implementation. His focus areas include migrating monolithic systems to scalable microservice architectures, launching new Java-based cloud projects, and building intelligent agentic workflows using LLM APIs (Anthropic, OpenAI, Ollama, llama.cpp).

## Contact

- Email: tom.seidel@remus-software.org
- LinkedIn: [linkedin.com/in/tomseidel](https://linkedin.com/in/tomseidel)
- GitHub: [github.com/tmseidel](https://github.com/tmseidel)

## Sections

- [Home](https://remus-software.org/): Overview, about, tech stack, and availability
- [Projects](https://remus-software.org/projects/): A curated selection of 30+ project references across automotive, medical technology, telecommunications, and more
- [Articles](https://remus-software.org/articles/): Technical articles on Java, cloud-native architecture, DevOps, and software engineering
- [Atom Feed](https://remus-software.org/feed.xml): Article feed in Atom format

## Articles


- [AI-Git-Bot vs Copilot, GitLab Duo, Qodo and OpenHands — Which Problem Are You Actually Trying To Solve?](https://remus-software.org/articles/ai-git-bot-vs-copilot-gitlab-duo-qodo-openhands/): An honest comparison of AI-Git-Bot, GitHub Copilot, GitLab Duo, Qodo and OpenHands. Not by features, but by the engineering problems they were built to solve.

- [We Gave Every Team Member an AI Agent — Here's What It Took to Make Them All Think Alike](https://remus-software.org/articles/howto-share-your-ai-experience/): Centralizing AI agent configuration across a team using Hermes Agent profiles and Git — keeping everyone on the same models, skills, and behavior without locking them into a web interface.

- [Why Your Local LLM Agent Fails at Round 4: The Hidden Context Window Crisis](https://remus-software.org/articles/context-window-hardening/): Agentic AI systems crash after 4-6 rounds when context windows fill up. A four-layer defense strategy keeps local agents running reliably for 20+ rounds with minimal accuracy loss.

- [From Qwen3-32B to Qwen3.6-35B-A3B: Upgrading a Local Inference Stack on 2× RTX 5060 Ti](https://remus-software.org/articles/qwen3-to-qwen36-upgrade/): A dense-to-sparse model swap that nearly doubled token throughput, improved output quality, and taught us a few things about vLLM with MoE and Mamba architectures.

- [I Built My Own AI Server — Here's What I Learned About Running Frontier Models at Home](https://remus-software.org/articles/i-built-my-own-ai-server/): A hands-on journey of building a local AI server with dual RTX 5060 Ti GPUs to run 32-billion-parameter models at home — covering hardware choices, software stack (Ollama, vLLM, Open WebUI), Blackwell compatibility headaches, and how local open-source models compare to frontier APIs.

- [Sovereign AI Is Loud — But The Real Issue Is Sovereign IT](https://remus-software.org/articles/regarding-sovereign-it/): Why the debate about sovereign AI is really a broader discussion about sovereign IT — control over infrastructure, platforms, portability, and the long-term ability to make and reverse technical decisions.

- [Meet AI-Git-Bot 1.7 — the teammate that reviews, tests and ships your PRs](https://remus-software.org/articles/meet-ai-git-bot-the-teammate-that-review-ships-test-prs/): AI-Git-Bot 1.7 is here, transforming from 'the PR review bot' to the AI teammate your repo has been waiting for — reviewing your code, writing your tests, deploying your previews, cleaning up after itself. The chores that always get cut under deadline pressure? Wire one bot. They get done. Every PR. Forever.

- [Standardizing Agentic AI at Scale: Why Remote MCP Servers Are the Missing Layer](https://remus-software.org/articles/remote-mcp-standardization/): Why organizations should standardize agentic AI with remote MCP server pools to improve consistency, governance, security, and adoption across teams.

- [Improving Software Quality with Reviewer Personas in AI-Git-Bot](https://remus-software.org/articles/improve-software-quality-with-reviewer-personas/): How to use AI-Git-Bot as a review gateway with multiple specialized reviewer personas to catch security, reliability, performance, testability, and maintainability issues in Git pull requests.

- [Running Docker Engine in WSL2 alongside Docker Desktop for Windows](https://remus-software.org/articles/kb_two_docker/): A practical guide to running both Docker Engine natively in WSL2 and Docker Desktop for Windows in complete isolation — diagnosing conflicts, fixing networking issues, and configuring coexistence.

- [Building an AI Agent for Code Generation: Lessons from 13 Iterations](https://remus-software.org/articles/how-to-develop-an-ai-agent/): Practical lessons learned from developing an agentic AI system that generates source code — from naive prompting to a robust, tool-using agent with a generic loop, schema-validated plans, provider-native function calls, and an optional self-critique step.

- [Restic Explorer 1.0 — A Lightweight Monitoring Dashboard for Restic Backups](https://remus-software.org/articles/rest-explorer-1-0-released/): Restic Explorer 1.0 is out — a lightweight, self-hosted web dashboard that monitors all restic backup repositories across S3, Azure, SFTP, REST, and Rclone from a single UI with automated scans, integrity checks, and retention policy tracking.

- [From Legacy to Lean: Rethinking Your Backup Strategy](https://remus-software.org/articles/replacing-veeam-with-restic/): How we replaced a costly, complex backup system with a simple shell script and S3 storage — and the key questions to ask before you do the same.

- [Evaluating Self-Hosted AI Services: A Translation Service Case Study](https://remus-software.org/articles/self-hosted-ai-translation-service/): A practical evaluation of replacing DeepL with a self-hosted translation service using open-source LLMs — comparing quality, performance, and cost.

- [Migrating a Monolith to Microservices: A Practical Guide](https://remus-software.org/articles/monolith-to-microservices/): A hands-on walkthrough of the architectural decisions and patterns I use when migrating Java monoliths to cloud-native microservices.


## Skills & Technologies

Languages: Java, Kotlin, Bash, Python, TypeScript
AI & LLM: Anthropic API, OpenAI API, Ollama, llama.cpp, Agentic Workflows, Hugging Face
Frameworks: Spring Boot, Micronaut, OSGi, React
Cloud Platforms: AWS, Azure, OVH, Hetzner
DevOps: Docker, Kubernetes, Helm, Terraform, Maven, Ansible
CI/CD: GitHub Actions, Jenkins, GitLab CI
Tools & Protocols: Git, REST, gRPC, MQTT, WebSockets, oAuth2

## Availability

Open for freelance and contract work. Remote-first, based in Germany (CET/CEST timezone). Prefers remote work but open to on-site at regular intervals.

## Optional

For the full-text version of all content, see: https://remus-software.org/llms-full.txt