- Research, design and optimize AI Agent architecture: planning, tool-use / function calling, memory, multi-agent orchestration for RAGEngine and GreenNode's AI products.
- Build and optimize end-to-end Agentic RAG pipeline: ingestion, chunking, embedding, hybrid retrieval (dense/sparse), reranking, query rewriting/decomposition, iterative & self-reflective retrieval.
- Design agent and RAG quality evaluation system: faithfulness, answer relevancy, retrieval precision/recall; build benchmark and automated evaluation process (LLM-as-a-judge, RAGAS or equivalent).
- Optimize simultaneously quality – latency – cost: semantic caching, model routing, prompt optimization, context management/compression.
- Coordinate with product and infrastructure teams to bring research results into production on GreenNode's self-host model platform.
- Monitor research and new standards (agent frameworks, MCP, A2A, deep research patterns...) and propose application to products.
Requirements
- From 2 years (Middle) to 4+ years (Senior) AI/ML Engineering experience, of which minimum 1–2 years working directly with LLM applications.
- Proficient in Python; experience with agent/RAG frameworks (LangChain/LangGraph, LlamaIndex, AutoGen, CrewAI...) or building agent systems from scratch.
- Deep understanding of LLM in practice: prompt engineering, function calling, structured output, context window management.
- Experience with embedding and vector database (Milvus, Qdrant, Weaviate, pgvector...), search/IR techniques: BM25, hybrid search, reranker (cross-encoder).
- Deployed LLM systems to production: API design, monitoring, observability/tracing (Langfuse, OpenTelemetry or equivalent).
- Good English reading comprehension for technical documentation and papers.
