Unleashing the Power of Unstructured Data
summary
Holmes starts as a proof of concept and becomes a global real time search engine for ING wholesale banking that indexes text PDFs web pages and credit packs and delivers results in under two minutes while keeping originals out of the index. Architecture uses event driven microservices Kafka S3 Elasticsearch OCR table extraction is the hardest task Postgres then ClickHouse for analytics FinBERT and NER power sector classification and recommendations which surface relevant counterparties unexpectedly Holmes: PoC → global real-time search engine reduced search-to-insight time under two minutes event-driven Kafka + Elasticsearch + S3
AI-written summary · about
chapters (6)
- 0:00Unstructured data search
- 3:20Internal search engine need
- 8:01Search engine architecture
- 16:17Technical stack & design
- 24:28Dual-track development
- 31:35Summary & Q&A
related talks
22mEvolving Search Architecture for a Demanding Tesco Online Business
49mReplatforming Online Search for a Major High Street Retailer
49mTransforming Analytical Data with Event-Driven Architecture
32mBusiness Intelligence in Microservice architecture #bbuzz
40mA Developer’s Adventures in AI Wonderland
44mThe Art and Benefits of Computing Eventfully
43mOpenSearch & AI Driven Cloud Native Platforms: Adaptive & Agentic Microserv...
15mDealing with Bad Architecture Decisions: A Case Study