I help engineering teams ship with AI agents, without losing control of their code.
Specs, guardrails and evals, set up in your own repository. Ten years of building and leading product teams, now focused on making agent-written software hold up in production.
Available now for a senior role or contract work. Based in Vienna, working remotely across Europe.
Worked with
Three ways I can help your team.
Pick the one that fits where your team is today. Every engagement runs the same way.
-
Assess
When you want to know where your team stands.
How your team uses AI today
A short, concrete review of how your engineers work with coding agents: where it helps, where it quietly erodes quality, and a prioritised plan for what to change first.
A short, fixed-scope review
-
Enable
When you want your team working this way.
Workshop and setup, in your repository
One or two days with your team on your own codebase: a loop from intent to shipped code, specs and decision records as the contract with agents, guardrails agents cannot bypass, and evals for the AI features you ship.
One or two days, on your codebase
-
Build
When you need AI features designed and shipped.
Fractional AI product engineering
Hands-on design and delivery of AI features in your product: LLM pipelines with structured outputs, evaluation and observability, built with EU AI regulation in mind.
One to two days a week, embedded
- DiagnoseI read the code and watch how the team actually works, rather than trusting the documentation about it.
- ProveChange one workflow end to end and measure it, before rolling anything out more widely.
- Hand overEverything lives in your repository, and your team runs it without me.
Hiring rather than contracting? I am also open to a senior engineering, technical lead or AI engineering role. Talk about a role
What I work with.
Retrieval, evals and guardrails, built on the model providers' SDKs and RubyLLM, agents over MCP, and ten years of Ruby, Rails and PostgreSQL underneath.
- What I build
- RAG
- LLM pipelines
- Evals
- Structured outputs
- Spec-driven workflows
- Agent guardrails
- Harness extensions
- Prompt-injection screening
- AI and agents
Claude
OpenAI
Gemini
Kimi- RubyLLM
MCP- Langfuse
- pgvector
- Languages
Ruby
Go
Rust
TypeScript
Python
JavaScript
C++
C
- Frameworks and data
Rails
Sidekiq
Ember
React
Vue
Svelte
PostgreSQL
MySQL
DuckDB
- Infrastructure
Docker
Kubernetes
Helm
Google Cloud- Kamal
I build the guardrails I teach.
The tools behind my workshops are ones I build or help build. Most are open source, so you can read how they work before you bring me in.
-
enola
Architectural regression testing for AI-assisted development. I built the constraints program: architectural laws declared over the code graph, checked deterministically in CI, with a Ruby DSL and four gems for Rails teams.
-
pi-brain
An intent store for agentic development on the pi harness: human-gated agent workflows, decision records agents must read before editing, and agent-written knowledge that cannot promote itself. Published on npm, nine tagged releases.
-
tabula
Shared memory for teams of agents and people, written in Rust: files and git as the truth, claims as leases so agents never collide, an append-only journal of who did what, and the decisions that belong to people refused by the tool itself.
What people I have worked with say.
-
Muhamed is one of the most amazing professionals I have ever worked with. He is extremely talented, dependable, and disciplined. … I would strongly recommend Muhamed and would love to work with him again.
Read the full recommendation
Muhamed is one of the most amazing professionals I have ever worked with. He is extremely talented, dependable, and disciplined. Moreover, he is always willing to push himself beyond limits to make a product successful. With Muhamed's excellent communication and engineering management skills everything is possible. At Experfy, Muhamed single-handedly managed the team collaboration consisting of 7+ timezones. He has a vast reservoir of talent - enabling him to lead Backend, Frontend, and DevOps teams with flying colors. In design thinking sessions, Muhamed played a vital role with his powerful insight about the product roadmap and technical knowledge. As an engineering manager, he always backed his team with full support. In addition to these Muhamed conducted extremely high-yielding Sprint retrospectives. He was always willing to go the extra mile to increase productivity and eliminate blockers. I would strongly recommend Muhamed and would love to work with him again.
Hassan AkbarWorked with Muhamed at Experfy, 2021
-
Muhamed leads by doing. When we hit a hard problem, he didn't start with a long discussion. He wrote the code, showed it working, and then we talked. That habit moved the team forward more than once.
Read the full recommendation
I worked with Muhamed on the same team for a year, and he quickly became one of the people I relied on most. Muhamed leads by doing. When we hit a hard problem, he didn't start with a long discussion. He wrote the code, showed it working, and then we talked. That habit moved the team forward more than once. He is a strong Ruby on Rails engineer and a natural architect. He took on leading the technical direction and shaping the code, and his decisions held up as the project grew. No challenge was too big or too unfamiliar for him. He is resourceful and knowledgeable, and when the answer wasn't obvious, he knew where to look. He is precise and professional, and he holds a high bar. He is also direct. When something wasn't good enough, he said so plainly, even when that was hard to hear. It was always about the work and never about the person, and it made both our code and our team better. I would gladly work with Muhamed again, and I recommend him to any team that needs a Rails engineer who can lead.
Zlatko AlomerovicWorked with Muhamed on the same team
- Founded and organised RubyConf Austria 2026 in Vienna, with a programme including Dave Thomas, Chad Fowler, José Valim and Armin Ronacher.
- Took full ownership of EuRuKo 2024 in Sarajevo, Europe's longest-running Ruby conference.
Working on something where AI has to hold up in production?
Tell me what you are building, and I will tell you honestly whether I can help.