Enterprise Search Engineer
Pubblicato il 13-08-2026 - Ntt Data Europe & Latam in Roma
NTT DATA is looking for an Enterprise Search Engineer (Elastic), to perform on remote or onsite from Rome for a UN Agency. You will work closely with ICT teams to deliver a well-engineered functioning enterprise search with Artificial Intelligence (AI) augmentation based on Enterprise Elastic search.
In this role you will be responsible for the following tasks:
- Design and implement scalable data ingestion pipelines and connectors to ingest structured, semi-structured, and unstructured content from enterprise sources (SharePoint, Liferay, Web crawls, Data Lake, Corporate Systems, etc.) into Elasticsearch or an equivalent search index, supporting batch, incremental, and near-real-time indexing processes.
- Design mechanisms to track document versions, source provenance, access permissions, timestamp updates, and deletion events to keep the search index accurate and current; develop content extraction pipelines for PDF, Word, Excel, PowerPoint, HTML, emails, scanned documents, and other enterprise formats,
converting them into standard markdown and/or vector embeddings to improve AI readability.
- Design and implement semantic chunking strategies and hybrid search logic optimized for retrieval quality (chunk size, overlap, section-aware splitting, heading preservation, table handling, context retention), and implement metadata extraction, enrichment, and deduplication of content during ingestion.
Build Retrieval Capabilities:
- Develop hybrid search capabilities combining keyword-based search, semantic vector search, metadata filtering, and contextual retrieval, complemented by re-ranking pipelines using specialized embedding models, ranking logic, or other suitable re-ranking techniques to improve relevance of retrieved results.
- Implement advanced retrieval techniques such as query rewriting, query expansion, function or tool calling, multi-query retrieval, metadata-aware retrieval, parent-child retrieval, contextual document embeddings, and contextual compressio