Senior/Lead AI/ML Performance Engineer

EPAM Systems Inc; United States; Remote; USD 150,000 - 230,000; Full time

Job summary

EPAM Systems is seeking a Senior/Lead AI/ML Performance Engineer to benchmark and profile AI/ML workloads on TPU hardware, generating empirical data for the optimization solver's cost model. You will conduct JAX/XLA benchmarking on TPU slices and analyze TPU profiler traces to build hardware coefficient matrices.

The role involves profiling LLM training and inference performance, collaborating with optimization engineers to calibrate solver inputs and improve accuracy and efficiency of cost

Qualifications

• Strong understanding of the modern AI/ML landscape and LLM architectures.
• Hands-on experience with LLM models/pipelines (training and inference).
• Proficiency in Python and/or Golang.
• Familiarity with TensorFlow, PyTorch, and LangChain.
• Experience with JAX/XLA and TPU-specific profiling tools.

Responsibilities

• Conduct JAX/XLA benchmarking on physical TPU slices.
• Capture and analyze TPU Profiler traces.
• Build hardware coefficient matrices for use in the optimizer's cost model.
• Profile LLM training/inference performance (FLOPS, memory access, token throughput).
• Collaborate with Optimization Engineers to calibrate solver inputs.

Skills

AI/ML landscape knowledge

LLM architectures

Python

Golang

Tools

TensorFlow

PyTorch

LangChain

JAX/XLA

TPU profiling tools

Job description

We are looking for a Senior/Lead AI/ML Performance Engineer to benchmark and profile AI/ML workloads on TPU hardware to generate the empirical performance data that feeds the optimization solver's cost model.

Responsibilities

• Conduct JAX/XLA benchmarking on physical TPU slices
• Capture and analyze TPU Profiler traces
• Build hardware coefficient matrices for use in the optimizer's cost model
• Profile LLM training/inference performance (FLOPS, memory access, token throughput)
• Collaborate with Optimization Engineers to calibrate solver inputs

Requirements

• Strong understanding of the modern AI/ML landscape and LLM architectures
• Hands-on experience with LLM models/pipelines (training and inference)
• Proficiency in Python and/or Golang
• Familiarity with TensorFlow, PyTorch, and LangChain
• Experience with JAX/XLA and TPU-specific profiling tools

Nice to have

• Experience with GPU/TPU performance benchmarking at scale
• Background in ML systems or MLPerf-style benchmarking