Description
We are the GigaChat Data team, preparing data for training the entire line of GigaChat LLM models (GigaChat, GigaChat Vision, GigaChat Audio, etc.). We specialize in creating and improving prompts (prompt-engineering), which will help us develop our products by enhancing the quality of interaction between models and users.
Responsibilities
Tasks:
- development and optimization of complex system and user prompts
- integration of LLMs via API into product scenarios
- testing prompts, conducting experiments and optimizing them based on quality metrics
- performing fact-checking and reviewing model responses, eliminating hallucinations and improving style
- analyzing datasets, identifying and eliminating bottlenecks
- working in tandem with AI trainers, ML engineers, analysts, and the product team.
Requirements
We expect from you:
- experience working with LLM APIs (GigaChat, OpenAI, Anthropic, local models, etc.)
- practical experience in creating and optimizing system and user prompts using various methods (few-shot, zero-shot, chain-of-thought, self-consistency, etc.), configuring generation parameters (temperature, top_p, max_tokens, etc.), and adapting prompts for different model usage scenarios
- experience in testing and evaluating prompts
- confident proficiency in Python for automation (pandas, requests/httpx, working with JSON, data processing)
- understanding of RAG and tool-calling principles – ability to design and implement pipelines with retrieval-augmented generation; knowledge of approaches to integrating tools (tool-calling) for executing external functions, working with APIs
- understanding of LLM operating principles, model training stages, and basic statistical metrics used in ML.
Will be a plus:
- experience with LangChain, GigaChain, or equivalents
- skill in writing prompt generation and validation pipelines
- knowledge of English at a technical documentation reading level.
Conditions
- remote work format for Kazan and the Republic of Tatarstan, including from cities without regional coefficients
- annual salary review and annual bonus
- more than 400 educational programs from SberUniversity for professional and career development
- extended voluntary health insurance (VHI), preferential family insurance, and a corporate pension program
- flexible mortgage discount equal to 1/3 of the Central Bank's key rate
- free SberPrime+ subscription, discounts on partner company products
- referral bonus for recommending friends to the Sber team.
