SENIOR HPC/COMPUTE ARCHITECT

Company: Crear Group
Apply for the SENIOR HPC/COMPUTE ARCHITECT
Location: London
Job Description:

SENIOR HPC/COMPUTE ARCHITECT

ROLE OVERVIEW

Owns the strategy and design for optimising risk computation stack across both the C++ engine and the Rust library, from initial architecture through production stabilisation and intraday extension. Responsible for architecture, validation strategy, and technical metrics framework.

KEY RESPONSIBILITIES

  • Define the performance optimisation architecture for the computation stack – C++ library and Rust library paths, with architecture diagrams, sequenced delivery plan, and explicit EOD and near-Real Time performance targets
  • Establish the test and validation strategy: benchmark definitions, regression test suites, acceptance criteria per milestone
  • Define the technical metrics framework: what constitutes an optimisation gain, how it is measured
  • Direct C++/Rust Compute Engineers Real Time and intra-day compute batch PoC execution
  • Throughout: Provide Technical Lead with compute-side interface specifications for quant library integration

TECHNICAL REQUIREMENTS – MANDATORY

  • Expert C++ (8+ years in production) in numerical computation or financial risk engine context – SIMD, cache efficiency, memory layout for numerical workloads
  • Production Rust (3+ years): memory management, async runtime (Tokio), unsafe Rust, FFI; profiling and optimising compute workloads end-to-end
  • Grid/HPC architecture: distributed computation orchestration, job scheduling, workload partitioning for large-scale risk batch; GCP-native grid or cloud HPC frameworks
  • GCP compute infrastructure: GKE batch workloads, GCE instance selection for compute-intensive workloads, Cloud Batch or equivalent
  • Performance profiling discipline: CPU profilers (perf, VTune, gperf), memory allocator analysis, latency measurement and statistical benchmarking methodology
  • Near-Real Time architecture: sub-second to seconds latency compute systems – different scheduling, state management, and compute allocation vs EOD batch
  • C++/Rust interoperability: FFI, gRPC, or proto-based bridging to Java layer

DOMAIN CONTEXT – REQUIRED

  • Investment bank risk computation experience at engine level: pricing engines, sensitivity computation, scenario-based risk frameworks inside a tier-one institution or equivalent
  • Equity derivatives product familiarity: options, warrants, equity finance structures

#J-18808-Ljbffr…

Posted: June 15th, 2026