Work format: remote (no office) City: doesn't matter, we work across the Russian Federation; comfortable to overlap with Moscow time (MSK ±3 hours) Employment: project-based, with the prospect of transitioning to full-time cooperation Salary range: from 180,000 to 250,000 ₽/month (net)

Job description: We are looking for a Backend / ML Infrastructure / MLOps Engineer for a project — an AI assistant for Russian judicial practice (SaaS for lawyers). We are building a product with a clear architecture: we collect RAG from the array of Russian court documents and connect it to a self-hosted LLM with tools. The user formulates a query in plain language and receives a detailed answer with links to specific cases. The prototype is already working, parsing and indexing are in progress — the task is to finalize the product.

Tech Stack: Python and TypeScript, Postgres, Qdrant (vector search), embeddings, RAG + Reranking (Qwen), self-host LLM inference on own GPUs, S3 (Backblaze B2), Sentry.

What's already done: — Semantic search of judicial practice, selection of similar cases, analysis of court positions, summaries of specific cases. Partially done: analytics on instances and judges; links to specific cases. What needs to be completed (area of work): — Statistical tools for profiling actors — Finalize infrastructure — Generation of procedural documents (claims, appeals, complaints) with export to .docx/.pdf — Productionizing RAG + reranking and stable self-host LLM inference — API, monitoring and logging of response quality, load stability

Who we are looking for: middle / junior-to-middle, confident in Python (TypeScript is a plus). The main thing is the ability to understand existing code and bring features to production. Experience with RAG / LLM / vector search / parsing large text arrays is a big plus. There is a detailed backlog and a lead developer who will conduct the interview and onboard you to the tasks.

Company name: OOO «Образовательные продукты Русяева»

Contacts: Andrey, recruiter contact