PULSAR: Pooled Unified Late-Interaction Search and Retrieval for Enterprise Visual Document RAG
2026-08-28 • Information Retrieval
Information Retrieval
AI summaryⓘ
The authors developed PULSAR, a visual search system used by an investment company to quickly find information in large sets of images like pitch decks and reports. Instead of converting images to text (OCR), PULSAR indexes images directly using a special neural network, making searches much faster and cheaper. Their method reduces search time significantly with minimal loss in accuracy and handles many searches at once better than older systems. Since its launch, PULSAR has processed millions of pages and improved answer accuracy compared to previous OCR-based tools.
vector searchOCR (Optical Character Recognition)image indexingneural network backboneMaxSim rescoringquery per second (QPS)NDCG (Normalized Discounted Cumulative Gain)Recall@10pitch decksdiligence materials
Authors
Benjamin Constable, Anup Roy, Vishal Sharma, Rishabh Upadhyay, Robin Mills, Aidan Millar
Abstract
Institutional investors search visually dense pitch decks, board packs, and diligence materials that change hourly near deal closing. OCR followed by figure verbalisation is costly to refresh at this scale and can lose chart detail. We present PULSAR, a production vision-first retrieval system deployed at Mubadala Investment Company. PULSAR indexes page images with a frozen ColPali-style backbone and uses a pooled two-stage late-interaction index: compact page summaries support initial retrieval, followed by exact MaxSim rescoring over a finer pooled representation. On ViDoRe V3, this design reduces median vector-search latency by 15.1 times against an unpooled configuration with less than 0.01 absolute NDCG@10 and Recall@10 loss; production median vector-search latency is 156 ms. Under concurrent load, the pooled index sustains approximately 88 times higher QPS than an unpooled index. The event-driven ingestion path is estimated to be approximately 20 times cheaper per page than the OCR+verbalisation baseline it replaced. Since March 2026, PULSAR has served 78 thousand documents and approximately 2.4 million pages across more than 3,000 deals. At the production top K, it more than doubles answer-fact recall over the OCR+verbalisation baseline.