Real-Time Financial Risk Computation for a Global Investment Bank

A leading global investment bank was running its entire risk computation as an overnight batch process -- 4 to 6 hours to reprice the full portfolio. Traders started each day with yesterday's numbers in today's market. Missed hedging opportunities, delayed responses to market moves, and growing FRTB reporting gaps were costing the bank millions.

Euclid built QGraph, a reactive computation engine on a live dependency graph that delivers sub-millisecond pricing latency across 1M+ instruments. The platform replaced batch recomputation with event-driven propagation, achieving a 10x throughput increase while reducing infrastructure costs by 60%.

About the Client

The client is a leading global investment bank managing a vast derivatives portfolio encompassing interest rate swaps, equity futures, and cross-currency instruments. Their trading desks operate across multiple time zones, requiring continuous, real-time risk visibility.

With millions of instruments under management and increasingly stringent regulatory requirements -- particularly the Fundamental Review of the Trading Book (FRTB) -- the bank needed to fundamentally rethink how risk was computed and delivered.

The Challenge

4-6 hour batch jobs. T-1 risk data. Millions at stake.

The bank's legacy risk engine ran as an overnight batch process, taking 4 to 6 hours to reprice the full portfolio. This meant traders started each day with T-1 risk data -- yesterday's numbers in today's market.

The consequences were tangible: missed hedging opportunities worth millions, delayed responses to market moves, and growing gaps in FRTB regulatory reporting. Intraday risk was effectively invisible.

The existing grid compute infrastructure was expensive, rigid, and architecturally incapable of supporting the real-time, instrument-level granularity the business required.

Our Approach

QGraph: a reactive computation engine on a live dependency graph.

Phase 1: Live Dependency Graph Architecture

Modeled the entire instrument universe as a directed acyclic graph -- every swap, future, and cross-currency instrument represented as a node with explicit dependencies on market data, curves, and counterparty exposures.

Phase 2: Event-Driven Propagation Engine

Replaced batch recomputation with reactive event propagation: when a market tick or position change arrives, only the affected subgraph recalculates -- eliminating redundant work and achieving sub-millisecond latency.

Phase 3: Incremental Risk Computation

Built fine-grained pricing functions for interest rate swaps, equity futures, and cross-currency instruments that compute only the delta from the last known state, not the full portfolio from scratch.

Phase 4: Horizontal Scaling Layer

Designed partition-aware compute distribution across commodity hardware, enabling 1M+ instruments per cycle while reducing infrastructure costs by 60% versus the legacy grid compute model.

Phase 5: FRTB-Aligned Reporting Pipeline

Integrated QGraph output directly into regulatory reporting workflows, enabling same-day FRTB compliance -- replacing T-1 batch data that had created persistent reporting gaps.

Outcomes

From overnight batch to real-time risk:

• <1ms pricing latency -- real-time risk computation
• 1M+ instruments per cycle -- full portfolio coverage
• 10x throughput increase -- over legacy batch system
• $4M recovered in Q1 -- previously missed hedging
• 60% infrastructure cost reduction -- optimized compute footprint
• Same-day FRTB reporting -- regulatory compliance met

What Comes Next

With the core computation engine proven in production, the next phase extends QGraph's live dependency graph to front-office trading GUIs -- giving traders instrument-level risk in real time, directly within their existing workflows.

In parallel, the regulatory reporting pipeline is being expanded to feed live graph queries directly into compliance systems, replacing the last remaining batch-dependent workflows and establishing a fully continuous risk infrastructure -- from market tick to regulatory filing.