SRE - High-Performance Compute

Consultant - High-Performance Computing - Global Quant Firm

Contract (transition to perm) - London, UK - Competitive

We are seeking a highly skilled and motivated Consultant to join a leading quant technology firm specializing in leveraging innovative data science technology to deliver valuable insights and solutions.

The primary focus of this role is High-Performance Computing (HPC). Candidates should have strong expertise in UNIX/Linux systems, programming (Bash, Python, C++), GPU optimization, and high-performance storage solutions. Experience working in ultra-low-latency environments could be beneficial.

Key Responsibilities:

System Design and Optimization:

• Architect, deploy, and manage UNIX/Linux systems tailored for high-performance and high-frequency operations.
• Implement kernel-level enhancements, such as real-time or low-latency kernels, CPU pinning, and HugePages, to boost system performance.
• Optimize system settings to reduce latency, including IRQ balancing, process affinity, and memory management.

GPU Performance Tuning:

• Enhance GPU performance to accelerate quantitative models and simulations, ensuring high throughput and low latency for crucial financial computations.
• Conduct thorough tuning of GPU resources, focusing on memory management, parallel processing, and kernel optimization to maximize efficiency in high-frequency trading and complex data analysis.

HPC Programming and Optimization:

• Develop and optimize system-level code using C++ and other languages to support HPC needs.
• Apply advanced compiler optimizations and profiling tools, such as Intel VTune and perf, to identify and resolve performance issues.

Networking and Infrastructure:

• Employ high-performance, low-latency network interfaces like Solarflare or Mellanox and apply kernel bypass techniques using DPDK or PF_RING.
• Maintain precise time synchronization using Time-Sensitive Networking (TSN) and protocols like Precision Time Protocol (PTP) and GPS-based NTP servers.
• Utilize and manage network monitoring tools like Corvil to track and reduce latency.