The vacancy is well-structured with clear tasks and tech stack, but lacks company transparency.
Job description
We are creating an AI assistant that helps researchers quickly find, extract, and analyze data from various sources — from international statistical databases to large administrative datasets. The project is in the active prototyping stage: we need to conduct a series of experiments quickly, select working solutions, and assemble them into a production-ready service.
Responsibilities
### What You Will Do
- Agent architecture (orchestration, tool use, multi-step reasoning)
- Integrations with statistical bureau APIs, SQL generation, linking with RAG
- RAG components: searching metadata, API documentation, variables
- LLM through Yandex AI Studio (model selection, prompts, quality assessment)
- Vector search in Qdrant
- Data: PostgreSQL, DuckDB, Parquet in S3
- Experiments "from idea to prototype in 2-3 days"
Requirements
### What Is Required
- Understanding of agent systems and debugging
- Experience with RAG from prototype to production (hallucinations, quality assessment)
- Proficient in Python
- DuckDB / analytical DBs is a plus
- Comfort with uncertainty and independent decision-making
- English (read documentation, write prompts)
Conditions
### What We Offer
- Project-based employment in the staff or annual contract
About ООО "ОДСС"
OOO 'ODSS' (Open Data Systems & Solutions) specializes in software development, data engineering, collection, processing, analysis of data, digital archiving, and custom development. They focus on open data technologies, creating information portals for government, public organizations, and citizens, and conduct research on government and societal openness. Primary activity is computer software development, with tools like CKAN and Omeka S.