I will build ai automation and a custom ai agent in python
Über diesen Service
Most AI automations work on the demo and quietly die in week two.
I build the kind that does not. Mine has run unattended since 2016, rewritten from scratch four times: it pulls from sources, calls an LLM to summarise, tag and illustrate, then publishes - about 1,150 items a month. I also keep my own fork of an open-source runtime that runs AI agents in isolated containers: 115 commits of mine on routing, tool handlers with typed errors and retry, and recovery when a provider is overloaded.
What I build for you:
- Agents that actually act: read your data, decide, call your API, write back
- LLM steps with a fallback chain, so a dead provider degrades one item instead of stopping everything
- Scheduled jobs, queues, retries, and state that survives a restart
- Deployment on your server under systemd, with logs and alerts
Before freelancing I spent 11 years as a Linux systems administrator, up to 200 production servers under a 24/7 SLA. The failure modes that kill automations are the ones I have been paged for at 3am.
Python, not a no-code toy you cannot debug. The source code is yours.
Tell me the process you want gone and I will say honestly whether it is worth automating.
Lerne Kirill M kennen
AI automation engineer, Python, LLM, backend, DevOps, bots
- AusGeorgien
- Mitglied seitAug. 2026
- ⌀ Antwortzeit1 Stunde
- Letzte Lieferung4 Tage
Sprachen
Russisch, Englisch
Mein Portfolio
FAQ
Do I need my own server?
Yes, a small VPS is enough, usually around 5 dollars a month. If you do not have one, tell me and I will say exactly what to get, then set it up and deploy on it for you.
What happens when an AI provider goes down?
Nothing stops. Each stage has its own chain of providers, so the next one takes over automatically. In the worst case a single item is handled a bit more simply, instead of your automation going silent.
Can I see it working before I order?
Yes. Message me and I will walk you through the architecture of the system I already run in production, and answer any question about how your case would work.
How is this different from a Zapier or n8n workflow?
For simple chains those are great, and I will tell you if that is all you need. I get called when the workflow hits a wall: custom logic, an API that needs real error handling, an LLM step that must not fail, or something that has to run reliably for months.

