vacancy("Инженер AI") => {
  salary,
  skills,
  apply,
};

Инженер по искусственному интеллекту

О роли

Design and implement LLM-powered features end-to-end — from prompt architecture and model selection through API integration and production deployment — with minimal supervision.
Own prompt engineering for production features: design, version, and systematically evaluate prompts across model updates and behavior regressions.
Integrate conversational and agentic AI capabilities into an existing application, owning the API layer, session management, and graceful degradation strategies.

Build and maintain RAG pipelines — including chunking strategy, embedding selection, vector store management, and retrieval evaluation — tuned for the application's domain.
Work across retrieval approaches (dense vector search, BM25 hybrid, re-ranking) and evaluate trade-offs for accuracy, latency, and cost.

Select and apply frameworks (LangChain, LlamaIndex, LangGraph, custom) based on real trade-offs in the context of the product — not hype.
Build with and extend MCP (Model Context Protocol) servers for tool integration, external service access, and structured agent communication.

Define and run LLM evaluation pipelines — automated metrics, human eval, regression suites — and act on results without waiting for direction.
Identify prompt regressions, retrieval quality issues, and latency problems early and drive resolution.

Collaborate with backend and frontend engineers as a peer, translating AI capabilities into clean service contracts and integration specs.
Identify architectural or data quality issues early and escalate when scope warrants.
Stay current with the LLM ecosystem and bring concrete, well-reasoned proposals for adopting techniques or tooling that address real product problems.
Contribute to technical documentation, internal best practices, and code reviews for junior team members.

BSc or MSc in Computer Science, Machine Learning, AI, or a related field.
At least 1–2 years of hands-on experience in LLM engineering — through industry, coursework, or substantive personal projects.
Solid understanding of transformer-based LLM architectures and how model behavior, context windows, and inference parameters affect output.
Practical experience building RAG pipelines: chunking, embedding models, vector stores (Pinecone, Weaviate, pgvector, Chroma), and retrieval evaluation.
Familiarity with agentic frameworks and orchestration patterns: tool use, memory systems, multi-step reasoning, and agent-to-agent communication.
Understanding of MCP (Model Context Protocol) for building interoperable tool integrations and structured agent workflows.
Experience with LLM tooling such as LangChain, LlamaIndex, LangGraph, or equivalent — with an ability to go beyond the framework when needed.
Awareness of prompt evaluation techniques: LLM-as-judge, embedding similarity, regression testing, and structured output validation.
Strong data preprocessing skills: regex, normalization, pipeline design, and working with messy real-world data.
Proficiency in Python, with exposure to REST API design and async patterns.
Familiarity with containerization (Docker) and cloud deployment on Azure.
Comfort working in a codebase with legacy components and the judgment to integrate cleanly without over-engineering.

Müraciət üçün (CV):

Design and implement LLM-powered features end-to-end — from prompt architecture and model selection through API integration and production deployment — with minimal supervision.
Own prompt engineering for production features: design, version, and systematically evaluate prompts across model updates and behavior regressions.
Integrate conversational and agentic AI capabilities into an existing application, owning the API layer, session management, and graceful degradation strategies.

Build and maintain RAG pipelines — including chunking strategy, embedding selection, vector store management, and retrieval evaluation — tuned for the application's domain.
Work across retrieval approaches (dense vector search, BM25 hybrid, re-ranking) and evaluate trade-offs for accuracy, latency, and cost.

Select and apply frameworks (LangChain, LlamaIndex, LangGraph, custom) based on real trade-offs in the context of the product — not hype.
Build with and extend MCP (Model Context Protocol) servers for tool integration, external service access, and structured agent communication.

Define and run LLM evaluation pipelines — automated metrics, human eval, regression suites — and act on results without waiting for direction.
Identify prompt regressions, retrieval quality issues, and latency problems early and drive resolution.

Collaborate with backend and frontend engineers as a peer, translating AI capabilities into clean service contracts and integration specs.
Identify architectural or data quality issues early and escalate when scope warrants.
Stay current with the LLM ecosystem and bring concrete, well-reasoned proposals for adopting techniques or tooling that address real product problems.
Contribute to technical documentation, internal best practices, and code reviews for junior team members.

BSc or MSc in Computer Science, Machine Learning, AI, or a related field.
At least 1–2 years of hands-on experience in LLM engineering — through industry, coursework, or substantive personal projects.
Solid understanding of transformer-based LLM architectures and how model behavior, context windows, and inference parameters affect output.
Practical experience building RAG pipelines: chunking, embedding models, vector stores (Pinecone, Weaviate, pgvector, Chroma), and retrieval evaluation.
Familiarity with agentic frameworks and orchestration patterns: tool use, memory systems, multi-step reasoning, and agent-to-agent communication.
Understanding of MCP (Model Context Protocol) for building interoperable tool integrations and structured agent workflows.
Experience with LLM tooling such as LangChain, LlamaIndex, LangGraph, or equivalent — with an ability to go beyond the framework when needed.
Awareness of prompt evaluation techniques: LLM-as-judge, embedding similarity, regression testing, and structured output validation.
Strong data preprocessing skills: regex, normalization, pipeline design, and working with messy real-world data.
Proficiency in Python, with exposure to REST API design and async patterns.
Familiarity with containerization (Docker) and cloud deployment on Azure.
Comfort working in a codebase with legacy components and the judgment to integrate cleanly without over-engineering.

Müraciət üçün (CV):

Локация

Bakı

Опыт

1–2 лет

Занятость

Полная занятость

Зарплата

Не указана

Опубликовано

12 августа 2026