Engineering Blog

Deep dives into vector search algorithms, distributed systems, and the future of Aspected.

Illustration of document metadata structure. A stack of documents with a magnifying glass represents content analysis, while a diagram shows metadata fields such as author, title, subtitle, description, keywords, and series connected as structured attributes derived from a document.

Mar 8, 2026•Aran Montero Salvadó & 3 others

Metadata Enrichment as a First-Class AI Capability

How AI-generated metadata is reshaping enterprise retrieval and turning enriched metadata into a core component of relevance.

Read article

A flowchart illustrating how retrieval failures lead to LLM hallucinations. The process starts with a User Query, followed by a Retrieval Layer failing due to incomplete context or misaligned ranking. This leads to the LLM Generation stage where extrapolation is forced, resulting in a “Hallucination” where the root cause is actually Retrieval Failure.

Feb 19, 2026•Aran Montero Salvadó & 3 others

It Wasn’t a Hallucination. It Was a Retrieval Failure.

An analysis of the root causes of hallucinations in RAG systems, and a relevance-first approach that reduces them by design.

Read article

A comparison flowchart of Pre Filter and Post Filter strategies showing how documents are narrowed down by metadata first then text embedding, versus text embedding first then metadata.

Jan 22, 2026•Mervin de Jong & 2 others

Introducing Aspects: A New Vector Search Paradigm

A unified vector search approach that embeds both meaning and metadata into a single representation, making relevance more natural, expressive, and efficient.

Read article