Senior ML QA Engineer: Benchmark & Validate AI Systems

EngineersOfAI; Bristol; On-site; GBP 60,000 - 80,000; Full time

Benefits offered by this job

Unlimited annual leave

Up to 5% matched pension

Health cash plan

Free food

On-site barista

Job summary

EngineersOfAI in Bristol seeks a skilled individual to validate the ML stack that translates accelerator hardware into dependable AI performance. In this role, you will test and benchmark complex systems to ensure they operate reliably before reaching customers.

Your expertise will help expose regressions and correctness issues across various ML frameworks and guide teams in resolving these challenges. The role requires close collaboration with software and hardware teams to ensure quality and efficiency in model execution.

Qualifications

• Strong experience in Machine Learning or ML-adjacent software engineering roles.
• A solid grasp of neural networks, training, inference, numerical precision and performance trade-offs.
• Hands-on experience with PyTorch, TensorFlow, JAX, Triton or similar ML frameworks and tools.

Responsibilities

• Test, benchmark and validate complex AI systems before customer delivery.
• Run open source models, build automated benchmarking pipelines.
• Create targeted tests for low-level ML behaviour like numerical precision and distributed execution.

Skills

Machine Learning experience

Neural networks understanding

PyTorch experience

Python skills

Debugging skills in Linux

Tools

TensorFlow

JAX

Triton

Job description

EngineersOfAI in Bristol seeks a skilled individual to validate the ML stack that translates accelerator hardware into dependable AI performance. In this role, you will test and benchmark complex systems to ensure they operate reliably before reaching customers.

Your expertise will help expose regressions and correctness issues across various ML frameworks and guide teams in resolving these challenges. The role requires close collaboration with software and hardware teams to ensure quality and efficiency in model execution.