Electronic Trading
Billions of Messages. Zero Room for Delay.
Organization:The Goldman Sachs Group, Inc.
Role: Vice President: Head of Global Equities Trading and Client Connectivity
Team size:60
Project overview: Led the re-engineering of Goldman Sachs' equities trading connectivity and sales platforms — driving latency from 2–5ms to 0.3ms, consolidating fragmented client systems, and building proprietary sales tools that generated an estimated $40MM in annual productivity gains across a $7B revenue business.
Mission Statement
Electronic trading demands best-in-class performance to win in a latency-sensitive market, resilience at massive scale to process hundreds of millions of daily trades, and agility to adapt quickly to evolving client requirements.
"Clients don’t wait twelve weeks for us to adapt — their strategies change overnight. If we can’t deliver new capabilities quickly, they’ll take their flow somewhere else."
Solution
Two problems, both urgent. First: a FIX engine with bottlenecks that were costing Goldman Sachs competitive positioning on every trade. Second: 12 overlapping client preference systems that made it impossible to serve clients consistently or move quickly when they needed changes.
"In our business, a few milliseconds can be the difference between winning a client's trade and losing it to a competitor. If we can't be the fastest, we can't be the first call our clients make."
The solution was a ground-up re-engineering of the connectivity stack targeting sub-millisecond latency, a unified client preference system capable of handling 300+ changes per month, and a compressed release cycle that could actually keep pace with client demand.
Technical Details
Directed a 60-person global team across two parallel workstreams — infrastructure performance and sales effectiveness — both supporting $7B in annual equities revenue.
Connectivity and trading stack:
• Reduced FIX message latency from 2–5ms to 0.3ms via C++ optimization, stack re-engineering, process consolidation, and VIP-based routing
• Unified 12 client preference systems into one, enabling 300+ client changes per month with full workflow, audit history, and regression testing
• Compressed release cycle from 12+ weeks to 2 weeks through standardized upgrades, reduced version sprawl, a forced upgrade cycle, and auto-generated regression suites
• Improved performance SLAs by benchmarking observed peak vs. theoretical message rates and eliminating unnecessary messaging overhead
• Strengthened security through environment separation controls
Sales intelligence — conceived and built independently to identify revenue opportunities:
• Proprietary sales lead generation system (React, Elasticsearch, Java, Node.js) leveraging trades, positions, and client interaction data
• GSMaps — a secure full-stack GIS solution built on OpenStreetMap with heat mapping and route optimization, protecting sensitive client location data
Result
• 0.3ms latency — a clear competitive advantage in client connectivity
• Doubled client change throughput while improving stability and time-to-market
• 6x faster release cadence with improved SLAs and stronger security controls
• Estimated $40MM annual increase in sales productivity
• Earned federal R&D tax credit for architectural innovation