DevEx Engineer - Accelerate AI Inference Workflows

Fractile; Greater London; On-site; GBP 85,000 - 140,000; Full time

Benefits offered by this job

Competitive salary

Equity & Ownership

Private Medical, Dental and Vision

Pension

Job summary

Fractile, founded to accelerate AI inference, seeks a Developer Experience engineer in London or Bristol. You will join the Software organisation to shape how customers interact with the Fractile inference accelerator and to build tooling that speeds and simplifies usage.

You will identify tooling opportunities, design and deliver practical solutions for an LLM engineer's workflow with minimal disruption, and collaborate across teams to improve performance and developer experience.

Qualifications

• Experience with inference frameworks such as JAX and/or PyTorch.
• Experience of software development in Python, Rust, C/C++, or a similar language.
• Experience monitoring and optimising code written in CUDA, ROCm, OpenCL, or similar.
• Experience monitoring and optimising LLM deployments with 50+ billion parameters.
• Experience working with stakeholders across multiple teams.

Responsibilities

• Shape how customers interact with the Fractile inference accelerator and develop tooling to improve usability and performance.
• Identify opportunities for tooling and deliver solutions that fit into an existing LLM engineer's workflow with minimal disruption.

Skills

JAX/PyTorch

Python/Rust/C/C++

CUDA/OpenCL/ROCm

LLM deployments

Stakeholder collaboration

Tools

JAX/PyTorch tooling

Job description

Fractile, founded to accelerate AI inference, seeks a Developer Experience engineer in London or Bristol. You will join the Software organisation to shape how customers interact with the Fractile inference accelerator and to build tooling that speeds and simplifies usage.

You will identify tooling opportunities, design and deliver practical solutions for an LLM engineer's workflow with minimal disruption, and collaborate across teams to improve performance and developer experience.