I implement AI agents that actually run in production.
From audit to production in 2–4 weeks. Not a POC — agents that actually run.
Your AI prototypes work. Scaling them doesn't.
- —
You've built agents that run for one person, on one workflow.
- —
But production means permissions, monitoring, multi-user, error handling.
- —
Your team experiments with prompts — nobody owns the system.
- —
Meanwhile, you're creating positions that didn't exist 12 months ago, and the first hire won't ship for 6 months.
Agents in production, not a POC that stays on a shelf.
Three ways to work together
30+ agents shipped to production. I run my own multi-agent stack daily — every orchestration pattern I deploy comes from real usage. Here's how I work:
Know exactly where to start
Diagnostic & Blueprint
2–5 days
Starting at €2k
THE PROBLEM
You know AI agents could transform your ops. But nobody has mapped which workflows are actually worth automating.
FOR WHO
Any company that wants to know where to start with AI agents.
WHAT I DO
Audit of automatable workflows. Stack recommendation: Dust vs Claude Code vs custom vs hybrid. Prioritized implementation plan.
DELIVERABLE
Scoping document + recommended architecture + use case prioritization
RESULT
Stop guessing. Get a prioritized roadmap with clear ROI per use case.
First agents live in 2–4 weeks
Production Setup
2–4 weeks
€5–10k
THE PROBLEM
The gap between 'it works on my laptop' and 'it runs reliably for the team' is where most agent projects die.
FOR WHO
Teams ready to put their first agents in production.
WHAT I DO
Implementation of 2–3 agents in production. Connection to existing tools, instrumentation, team training.
DELIVERABLE
Agents in production + documentation + training session
RESULT
Your first agents run reliably, your team knows how to supervise them, and you have a foundation to build on.
I handle scoping and architecture. Technical implementation is done with a senior dev partner.
From prototype to platform
Scale & Industrialize
2–3 months
€15–30k
THE PROBLEM
Your agents work, but they're monolithic — one person runs them, nobody else can, and adding a use case means starting from scratch.
FOR WHO
Companies with agent prototypes that need to scale.
WHAT I DO
Decomposing monolithic agents into composable skills. Multi-user with permissions. Self-improvement loops. End-to-end instrumentation.
DELIVERABLE
Production-grade multi-agent system, documented and monitored
RESULT
A production-grade agentic platform your team can extend without you.
For companies that already have working prototypes and want to industrialize.
Systems in production
Not slides. Systems that run, with real users and business-critical decisions.
CONTEXT
My own production AI stack — daily dogfooding
SYSTEM
2nd Brain (Claude Cowork + Code), persistent markdown memory + Git + MCP, 9 specialized agents (GEO Auditor, Content Optimizer, Chief of Staff, Product Manager...), Supabase coordination bus for human-agent handoff, Claude Agent SDK + cron for automated workflows, Linear + Notion + Todoist orchestrated by agents
RESULT
I don't just implement agents for clients — I run my own multi-agent system daily. Every pattern I deploy comes from real production usage.
CONTEXT
Multi-agent orchestration for bid management (Sinay.ai)
SYSTEM
~30 Dust agents + central orchestrator, 24/7 automated workflows, real-time monitoring
RESULT
Bid decisions automated with minimal supervision. Running for months.
CONTEXT
AI agent in production for end customers (logistics, Sinay.ai)
SYSTEM
'Logistic Anomaly Detection' agent via GPT API, guardrails, monitoring, decision logging
RESULT
In production, used daily by end customers to detect supply chain anomalies. Not a POC.
CONTEXT
Growth instrumentation for B2B SaaS (Safecube.ai)
SYSTEM
Event tracking, multi-touch attribution, real-time dashboards, continuous experimentation
RESULT
0→5,000 MAU in 24 months. Growth decisions driven by data.
How it works
No workshops. No strategy decks. Each step is a natural funnel into the next. Start with a diagnostic. Go further only if it makes sense.
Discovery call
20 min, free. I understand your context and identify if I can help. No pitch, no deck.
OUTPUT
Honest answer on whether I can help and how.
Diagnostic & Blueprint
Audit of automatable workflows. Stack recommendation. Actionable implementation plan.
OUTPUT
Scoping document + recommended architecture.
Build & Deploy
First 2–3 agents in production. Connected to your stack, with guardrails, monitoring, and team training.
OUTPUT
Agents running in production.
Scale
Decompose and industrialize. Multi-agent architecture, self-improvement loops, end-to-end instrumentation.
OUTPUT
Production-grade agentic system. (optional)
This is for you if...
You want agents in production, not a POC that stays on a shelf
You have agent prototypes that work but don't scale
You're hiring your first AI Agent Engineer and need someone now
You've deployed Dust or Claude but can't get production value without expert help
You're looking for prompts or a chatbot
You want AI hype, not systems that run
You need strategy without execution
Operator, not consultant
13+ years in digital acquisition. Head of Growth at Sinay.ai: deployed Dust.tt across the company, ~30 agents in production, multi-agent orchestration for bid management running 24/7. I dogfood daily — my own stack runs 9 specialized agents, a Supabase coordination bus, and Claude Agent SDK for automated workflows. Now at Legallais (1,500-person B2B company) as an AI lab in a real SMB context.
I don't just implement agents for clients. I run them. Every guardrail, every orchestration pattern, every human-agent handoff design comes from production usage.
The cost of shipping code is trending to zero. Every simple tool gets 15 clones in 48 hours. The moat isn't in the code anymore — it's in the orchestration patterns, the context injection, the human-agent handoff. I learned these by building my own stack and dogfooding it daily.
For implementation-heavy engagements (Tier 2–3), I work with a senior dev partner (Fyher) who handles the technical build.
Work together
I take 2–3 engagements per quarter alongside my day job. Start with a 20-min call to see if I can help.