Browse categories
English
$
USD
Connect to your AI
I will create and deploy infrastructure and pipelines for production ai agents
Slava
About this gig
Your agent works in a notebook or on one VM. Now it has to survive real users: long jobs, API limits, token bills, and the 3am question of why it answered wrong. I will move it to production and make it observable.
What you get:
- Your agent or RAG app deployed on Cloud Run, ECS or Kubernetes
- Vector database: Qdrant managed or self-hosted, with backups
- Queue and workers so long agent runs survive HTTP timeouts
- Tracing per request: prompt, tool calls, latency and token cost
- Secrets manager, least privilege IAM, private networking, rate limits on LLM calls
- Eval step in CI so a prompt change cannot ship if quality drops
- Terraform for all of it, plus a runbook
Why me: I build these systems, not only host them. Markdex is my open source RAG engine on Qdrant, Loki MCP Server is my observability tool for agents, StatusHarbor is my own SaaS in production. 16 years in backend and infra.
How it works:
- You send the repo and where it runs today
- I send a plan and a monthly cost estimate within 24 hours
- I deploy to staging first, you test, then we cut over
- Handover call and documentation
You bring a working app. Building the agent is my other gig, happy to combine them.
Get to know Slava
Slava
Senior Software Engineer Technical Lead
- FromAlbania
- Member sinceJan 2018
- Last delivery2 months
Languages
English
You get AI that actually ships: agents that answer calls and chats around the clock, RAG that gives correct answers from your own data and automated workflows that cut manual work for your team. With 16+ years in software as an engineer and tech lead, I help you skip the failed-demo phase. I design the simplest architecture that fits your goal and deliver production-ready code with tests, docs and a clean handover. Typical path: a short call, an architecture plan in days, a working system in weeks, not months.
My Portfolio
Other AI Development Services I Offer
FAQ
Does my app need to be in a specific language?
No. Python, TypeScript and Go are the usual ones. If it runs in a container I can deploy it. If it does not yet, containerizing it is part of the Basic package.
Which vector database do you use?
Qdrant by default, managed or self-hosted in your cloud. If you already run pgvector, Pinecone or Weaviate, I keep what you have unless there is a clear reason to move.
Can you self-host the model?
Only when it is worth it. For most products a hosted API with rate limits and caching is cheaper and simpler. If you have a hard requirement for self-hosting, say so in the first message and I will quote it separately.
What does tracing give me?
Every request shows the prompt, retrieved chunks, tool calls, latency and token cost. When a user reports a bad answer you open that exact run instead of guessing.
Who pays for cloud and LLM usage?
You do, in your own accounts. I work through an IAM role or service account you can revoke, and I send a cost estimate before anything is applied.
Can you also build or fix the agent?
Yes, that is my AI agents gig. If you need both, message me first and I will scope one order instead of two.
How do you handle security?
API keys live in your cloud secrets manager, never in the repo. I work through an IAM role or service account you can revoke at any time. Traces are scrubbed of PII before storage and tool calls run against an allowlist. A full penetration test or compliance audit is out of scope.
