Expert.ai
AI Engineer
NewOnsiteMidModena, ItalyAI/ML
Our Take
AI Engineer at Expert.ai in Modena, Italy building production-grade LLM systems using RAG and agentic architectures.
What you’ll do
- Design and implement RAG pipelines end-to-end including ingestion chunking strategies embedding models vector retrieval reranking and response generation
- Build agentic workflows using LangChain LangGraph LlamaIndex or custom orchestration tool use memory management and multi-step reasoning
- Integrate and prompt engineer LLMs GP Claude Sonnet/Opus Qwen for domain-specific tasks and contribute to fine-tuning efforts when needed
- Develop MCP servers and clients to standardize tool and context exposure across AI systems
- Maintain vector databases FAISS Pinecone Weaviate Qdrant pgvector and optimize retrieval quality
- Build evaluation pipelines to track hallucination rate retrieval precision latency and output consistency over time
- Write clean tested production ready Python code and contribute to code reviews
- Collaborate with senior engineers and clients to translate requirements into solid technical decisions
What they’re looking for
- 2 to 4 years of software engineering experience
- 1 to 2 years focused on LLM or applied AI systems
- At least one production RAG or agentic application under your belt
- Solid understanding of embeddings, transformer fundamentals, context management, and prompt design patterns
- Familiarity with Docker
- Able to work with autonomy
Skills & Focus Areas
- RAG
- LLM
- LangChain
- LangGraph
- LlamaIndex
- Vector Databases
- FAISS
- Pinecone
- Weaviate
- Qdrant
- pgvector
- Python
As posted by Expert.ai
About expert.ai
We build production-grade AI systems for enterprise clients. Our work focuses on Large Language Models, Retrieval-Augmented Generation, agentic architectures, and knowledge-driven AI. We are a lean team of engineers and researchers who move fast and care about the quality of what we ship.
The Role
We are looking for an AI Engineer with around 2 to 4 years of experience — someone past the learning phase, who has shipped at least one LLM-powered system to production and knows what breaks, what scales, and what was a bad idea in hindsight.
You do not need to have done everything. You need to have done some things well, understand why they worked, and be ready to go deeper.
What You Will Do
- Design and implement RAG pipelines end-to-end: ingestion, chunking strategies, embedding models, vector retrieval, reranking, and response generation
- Build agentic workflows using LangChain, LangGraph, LlamaIndex, or custom orchestration — including tool use, memory management, and multi-step reasoning
- Integrate and prompt-engineer LLMs (GP Claude Sonnet/Opus, Qwen) for domain-specific tasks; contribute to fine-tuning efforts when needed
- Develop MCP servers and clients to standardize tool and context exposure across AI systems
- Maintain vector databases (FAISS, Pinecone, Weaviate, Qdrant, pgvector) and optimize retrieval quality
- Build evaluation pipelines to track hallucination rate, retrieval precision, latency, and output consistency over time
- Write clean, tested, production-ready Python and contribute to code reviews
- Collaborate with senior engineers and clients to translate requirements into solid technical decisions
What We Expect
- 2 to 4 years of software engineering experience, with at least 1 to 2 years focused on LLM or applied AI systems
- At least one production RAG or agentic application under your belt — you know what it took to get it there
- Solid understanding of embeddings, transformer fundamentals, context management, and prompt design patterns
- Familiarity with Docker
- Able to work with autonomy — you ask good questions, but you do not wait to be told what to do next
Formal degrees are welcome but not required. What matters is what you have shipped.
Nice to Have
- Experience with MCP protocols in real projects
- Knowledge graphs or GraphRAG pipelines (Neo4j, Neptune, or similar)
- Inference optimization: quantization (GGUF, AWQ, GPTQ), vLLM
- Evaluation tooling: RAGAS, TruLens, or custom eval design
- Domain NLP experience in legal, manifacturing, or healthcare