AI Inference Engineer

At F5, we strive to bring a better digital world to life. Our teams empower organizations across the globe to create, secure, and run applications that enhance how we experience our evolving digital world. We are passionate about cybersecurity, from protecting consumers from fraud to enabling companies to focus on innovation. Everything we do centers around people. That means we obsess over how to make the lives of our customers, and their customers, better.

And it means we prioritize a diverse F5 community where each individual can thrive.

The AI Inference Engineer plays a critical role in the AI lifecycle by bridging the gap between high-performance model development and optimized deployment environments. This position focuses on optimizing Large Language Models (LLMs) for inference, serving diverse environments—from GPU-rich data centers to resource-constrained edge devices with a strong emphasis on maximizing throughput, minimizing latency, and maintaining model accuracy. This role is pivotal in advancing F5’s AI capabilities, ensuring enterprise-grade reliability by leveraging hardware acceleration, designing scalable infrastructure, and monitoring system performance.

• High-Performance AI Serving Build and maintain robust inference engines using tools like vLLM, TGI (Text Generation Inference), and NVIDIA Triton, ensuring high performance at scale.
• Handle deployment optimizations to deliver low-latency AI serving solutions for multiple business applications.
• Hardware Acceleration and Optimization Profile and optimize models for specialized hardware backends, including NVIDIA GPUs (CUDA/TensorRT), Apple Silicon (CoreML), and AI accelerators like TPUs and LPUs. Collaborate with hardware teams to maximize utilization and performance across various computational environments.
• Inference Orchestration and Scalability Design and implement auto-scaling architectures for online (real-time) and batch inference pipelines, leveraging Kubernetes for inference routing and orchestration. Ensure software solutions are optimized for peak performance during traffic spikes, maintaining reliability and scalability.
• Performance Monitoring and Observability Establish robust observability frameworks to monitor Time to First Token (TTFT), tokens per second, and memory bandwidth utilization against service-level agreements (SLAs). Build and execute performance and load testing suites to identify bottlenecks and ensure consistent reliability at scale.

• Programming Languages
  :
  Proficiency in programming languages such as Python, C++, Rust, or Golang specifically for high-performance AI workflows.
• Inference Tools
  :
  Proven hands-on experience with tools like vLLM, TensorRT, Llama.cpp, and Ollama for inference development and optimization.
• Infrastructure Expertise
  :
  Strong familiarity with infrastructure technologies, including Docker, Kubernetes, and cloud platforms such as AWS, GCP, and Azure.
• Hardware Optimization Expertise
  :
  Comprehensive understanding of GPU and AI hardware, including techniques for profiling and optimizing performance for accelerators like NVIDIA GPUs and TPUs.
• Prior experience deploying LLMs
  :
  Prior experience deploying Large Language Models (LLMs) with advanced techniques like Speculative Decoding or Paged Attention.
• Contributions to open-source inference libraries or hardware-level kernel development
  :
  Contributions to open-source inference libraries or hardware-level kernel development (e.g., CUDA, Triton kernels).
• Background in MLOps or SRE roles
  :
  Background in MLOps or SRE roles focused on high-performance AI endpoints and reliability during demand surges.
• Proficiency in designing scalable solutions for high-throughput inference environments
  :
  Proficiency in designing scalable solutions for high-throughput inference environments optimized for traffic bursts.

• Latency Reduction
  :
  Continuously improve inference latency metrics, ensuring minimal Time to First Token (TTFT) and maximum tokens per second.
• Cost Efficiency
  :
  Achieve lower "Cost per 1K Tokens" through better resource utilization and hardware optimization.
• Scalability
  :
  Maintain…