Real-Time Search & Indexing

Real-Time Search & Event-Driven Listing Intelligence Platform

I architected and delivered a high-performance, real-time search and indexing platform using Elasticsearch and an event-driven architecture (EDA) to serve the Global Core. This system functioned as the authoritative search backbone for both consumer-facing applications and internal agent tools, powering the company’s global digital listing experience.

The platform was designed to ingest, normalize, and index listing data from hundreds of heterogeneous MLS feeds, each with unique schemas, update frequencies, and data quality constraints, while delivering sub-millisecond query performance at global scale.

Event-Driven Ingestion & Data Normalization

To handle continuous, high-volume listing updates, the ingestion pipeline was built around event-driven patterns: • MLS feeds published change events (create, update, delete) into the ingestion pipeline • Events were processed asynchronously to decouple ingestion from indexing • Normalization services transformed disparate MLS schemas into a canonical Global Core listing model • Validation and enrichment stages ensured data consistency, completeness, and compliance

This design enabled near real-time propagation of listing changes without introducing backpressure or system coupling.

Search Index Architecture & Query Optimization

Elasticsearch clusters were architected for low-latency, high-throughput search workloads: • Custom index mappings optimized for listing attributes, pricing, media, and metadata • Geospatial indexing supporting radius, polygon, and bounding-box queries • Dynamic attribute filtering across hundreds of listing facets • Query optimization strategies to support high-cardinality fields without performance degradation • Read-heavy scaling with shard allocation tuned for peak consumer traffic

Index refresh strategies balanced freshness with query performance during high update volumes.

Localization, Internationalization & Relevance

To support a truly global footprint, the search platform incorporated: • Multi-language indexing and analyzers for localized text search • Locale-aware sorting and relevance scoring • Currency and measurement normalization at query time • Brand- and region-specific relevance tuning via configurable scoring models

This ensured consistent and relevant search experiences across markets, languages, and brands.

Performance, Resilience & Observability

The platform was designed for continuous availability and operational resilience: • Event replay and reindexing capabilities for recovery and schema evolution • Backpressure controls to protect clusters during ingestion spikes • Integrated observability tracking ingestion lag, query latency, and index health • Automated scaling policies to handle seasonal and campaign-driven traffic surges

These safeguards ensured the search experience remained fast and reliable under sustained global load.

Business & Platform Impact

This architecture: • Served as the foundation for all digital listing discovery • Enabled real-time visibility of listing changes across consumer and agent channels • Supported advanced search capabilities that increased engagement and conversion • Scaled across brands, regions, and regulatory environments without duplication

By combining event-driven ingestion with a purpose-built search architecture, the platform delivered real-time intelligence at global scale, becoming a critical differentiator in the company’s digital ecosystem.