We run AI systems across hundreds of retail locations, public buildings and schools — edge-deployed, high-throughput, live in production. The core is solid. Now we're building the automation layer on top of it: agents that connect our models to the real world and actually get things done.

What you'll do

Design and build AI agents that automate real workflows — document processing, data pipelines, client system integrations, image analysis

Connect local self-hosted models (vLLM, Ollama) to agentic pipelines via MCP and tool use

Own your work end to end — from architecture decisions to deployment

Collaborate with a broader engineering org of 15+ developers across full-stack, backend, and infrastructure — the AI layer plugs into real, actively developed products

Must have

Strong Python — you've built things in production, not just notebooks

Solid experience building AI agents and agentic pipelines

Familiarity with MCP, tool use patterns, local model inference

Model output validation — you build evals and checks, you don't ship blind

You keep up with the field and know what's actually worth using

Nice to have

PyTorch / TensorFlow — training, fine-tuning

High-throughput data processing: multiprocessing, shared memory, IPC

Model optimization: quantization, TensorRT, ONNX

Cloud (Azure or AWS), Terraform, DevOps

Location

Bucharest. On-site, one day remote per week.

Team

Small, focused AI team embedded in a larger engineering org — 15+ developers across full-stack and infrastructure, always shipping. No bureaucracy, real ownership, plenty of surface area to work on.