back_office_ops · logistics · workflow
Grab experiments with LLM-assisted vector similarity search for improved relevance on complex queries
Traditional vector similarity search struggles with nuanced and negation-based queries, returning semantically close but contextually wrong results, limiting its usefulness for complex search requirements.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Vector search shortlisting
A vector similarity search is performed on the dataset to obtain a shortlist of potential matches for the given query.
Tools used
FAISSGPT-4otext-embedding-ada-002
Outcome
The LLM-assisted approach showed a clear advantage over raw vector search for complex queries and was deployed internally at Grab for larger datasets with similar improvements in search result relevance.
What failed first
Raw vector similarity search failed to handle negations, returning dishes containing the explicitly excluded ingredients when queried for food with no fish or shrimp.
Grounding & classification
Source type: technical build writeup
20 fields verified against source quotes.
enterprise searchragknowledge basefailure mode describedmetric backednamed customerproduction runtime claimedsource backedtools describedworkflow describedfinancial serviceslogisticsaccuracy improvementtechnical build writeupback office opsrag answering