AI Performance Engineer

[About the Role]

We are seeking an AI Performance Engineer to enhance the performance, scalability, and reliability of our ML and LLM-based systems. You will work across AI infrastructure, system performance engineering, model-serving optimization, and production MLOps, ensuring that workloads—such as RAG pipelines, agent workflows, semantic search, and automation platforms—run efficiently at scale.

In this role, you will build benchmarking frameworks, profile inference pipelines, identify system bottlenecks, and establish performance metrics to guide model selection, deployment strategies, and cost optimization. You will work closely with AI engineers to continuously refine and improve our production AI stack.

[Responsibilities]

1. Design, build, and maintain performance benchmarking pipelines for LLMs, embedding models, RAG components, vector databases, and agent runtime workflows.
2. Profile CPU/GPU utilization, memory bandwidth, latency, and throughput to identify performance bottlenecks across the AI serving stack.
3. Develop tools to measure, visualize, and track system performance, latency, and cost across environments.
4. Collaborate with AI engineers to optimize end-to-end agent workflows, including retrieval speed, tool invocation latency, task planning, and execution efficiency.
5. Develop deployment, monitoring, and observability strategies that align with the DevOps pipeline to ensure reliable and scalable AI workloads.
6. Define and validate performance KPIs for LLM serving, retrieval systems, and multi-agent orchestration.
7. Drive continuous improvement in system reliability, throughput, and resource utilization across production AI products.
8. Contribute to capacity planning, hardware utilization strategies, and inference cost optimization.

[Minimum Qualifications]

1. Bachelor’s degree (or equivalent practical experience).
2. 3+ years of software development experience in at least one programming language (Python preferred).
3. Strong understanding of ML system architecture, including training/inference pipelines, caching, distributed execution, and data flows.
4. Experience with performance profiling tools, such as Nsight, perf, PyTorch Profiler, or TensorBoard Profiler.
5. Hands-on experience with cloud environments, Docker, Kubernetes, and container orchestration.
6. Ability to evaluate trade-offs involving latency, throughput, cost, and hardware resources.