Machine Learning

• IT
• Chișinău-Centru
• 2-5 ani

What you'll work on:

The internal AI platform - the place for internal agents, where people can find info faster and automate workflows.

The ML platform - the place where data scientists models are trained, versioned and run on schedule against large datasets.

What you'll do:

• Develop and maintain the platform's services;
• Integrate third-party APIs and internal systems, databases, object storage and model providers withauth, retries, timeouts and failure modes handled properly;
• Connect AI solutions with internal applications, databases, APIs, and enterprise knowledge sources;
• Write tests, review your colleagues' code, keep dependencies current and the services debuggable;
• Work with data scientists to turn their needs into pipelines and tooling they actually adopt;
• Help develop reusable components, agentic capabilities, and integration patterns that accelerate enterprise-wide AI adoption.

What you need:

• Strong Python - 2+ years of production code. You have owned a service or library other peopledepended on: HTTP APIs (FastAPI/Flask), async basics, error handling and retries, structured logging,dependency management, unit and integration tests;
• Hands-on experience building LLM applications - context windows and token cost, retrieval overdocuments vs. feeding full context, embeddings and chunking, knowledge bases, tool/function calling,and the judgement to tell whether a bad answer came from the model, the prompt or the retrievedcontent;
• Practical data-science foundations. You have trained models yourself, even if not professionally, andhave good foundations in them;
• SQL and a relational database from application code - schema and migrations, indexes,transactions, connection pooling;
• Observability - metrics, logs and traces with Prometheus, Grafana, the ELK stack and distributedtracing (e.g. OpenTelemetry, Jaeger/Tempo);
• Model monitoring in production - data drift, concept drift and model performance decay.
• Comfortable with Kubernetes and Helm - you can deploy, inspect and debug a workload; Depth iswelcome but it is not the bar; you will grow it here;
• Docker, CI/CD (Azure DevOps / GitLab / GitHub Actions), Linux, git;
• GitOps delivery (ArgoCD or Flux), or the willingness to pick it up quickly.