Rebuilding FYERS Symbol Search
Moved symbol lookup from the transactional database to an indexed search service with custom fuzzy logic.
- Search Architecture
- Performance Optimization
- Infrastructure Cost
- Meilisearch
Pavan Kumar Segu
Senior Software Engineer 1 at FYERS
From their time as
Software Engineer 1
FYERS • 2023 - 2025
Overview
Pavan rebuilt FYERS symbol search around Meilisearch and custom fuzzy matching. LinkedIn records latency falling to 30 milliseconds and infrastructure cost and resource use falling by 80%.
The Story
Pavan took on symbol search when each typed query was placing work on the transactional database. Search had limited typo handling, and response times made symbol discovery slow for traders.
He chose Meilisearch rather than adding more database readers. The architectural decision separated search from OLTP, using reverse indexing for a workload MySQL was not designed to serve efficiently.
Pavan also wrote custom fuzzy-search logic to correct mistyped symbol queries. This preserved a forgiving user experience while moving query traffic onto a purpose-built search layer.
The revised system reduced search latency to 30 milliseconds according to LinkedIn. It also reduced infrastructure cost and resource use by 80%, while running on a small EC2 footprint.
Ownership Snapshot
Broad role
Backend engineer responsible for redesigning symbol search at FYERS.
Goal
Make symbol lookup faster, typo-tolerant, and less expensive to operate.
Direct ownership
Selected the search architecture and implemented custom fuzzy matching.
Team execution
Worked within the FYERS engineering organization to deploy the revised search flow.