AI Engineer
Pubblicato il 15-08-2026 - Altro in Cologno Monzese
TXT Group is looking for an AI Engineer – LLM & RAG Systems to join our team dedicated to developing cutting‑edge Generative AI solutions.
The successful candidate will be involved in the design, development and evolution of Agentic AI and Retrieval‑Augmented Generation (RAG) systems, contributing to the delivery of reliable, scalable and cost‑effective Large Language Model (LLM) -based solutions that address real business needs.
Main Responsibilities
Design and implement end-to-end agentic systems, including orchestration, intent classification, request routing, and query rewriting and expansion;
Define and maintain prompt engineering strategies tailored to different use cases;
Select and optimise LLM and embedding models based on quality, latency and cost;
Develop ingestion pipelines for unstructured data, including document parsing, cleaning, normalisation, chunking and metadata extraction;
Implement embedding generation pipelines and manage vector databases;
Design advanced retrieval strategies, including RAG, hybrid search and reranking techniques;
Implement function/tool‑calling patterns and integrate external tools (databases, applications, Machine Learning models and web search) using the MCP protocol;
Develop entity and relationship extraction processes to build knowledge graphs that enhance retrieval capabilities;
Define evaluation metrics and frameworks for RAG and LLM systems, measuring retrieval quality, answer correctness and relevance, hallucination detection, and developing regression testing suites for prompts and pipelines;
Design mechanisms to collect and analyse user feedback, supporting offline evaluation and potential model fine‑tuning activities.
Required Technical Skills
At least 2 years of experience developing solutions based on Large Language Models;
Strong Python skills and proven experience developing end-to-end LLM‑based applications;
Experience with LLM and agent orchestration frameworks (such as LangChain, LlamaIndex or equivalent) an