EmergingAI compilers, kernels and runtime systems

AI Performance Portability Engineer

Also called ML Portability Engineer

An AI Performance Portability Engineer keeps models fast and accurate across different hardware. They cut dependence on one chip or vendor, which widens supply options and protects serving costs and customer commitments.

What they do

The work starts with a model that runs well on one accelerator and poorly, or not at all, on another. The engineer profiles both, finds where time and accuracy are lost, and decides what to fix. Sometimes that means rewriting a kernel in a portable language. Sometimes it means changing how the compiler lowers an operation, or adding a backend abstraction. They build benchmarks that run across devices, check numerical agreement, and track regressions as models and drivers change. They work closely with model researchers, kernel authors and infrastructure teams.

What a strong hire delivers

In the first months, a strong hire gets one important model running on a second hardware target with quality and speed within an agreed margin. They leave behind a cross-device benchmark the team trusts and can rerun. They document which operations are portable and which are not, and they give the team a clear view of the remaining gaps and what each would cost to close.

How you can tell they are good

Ask them to walk through a real port. Good candidates explain what they measured first, where the loss came from, and what they chose not to optimise. They talk about numerical differences, not just speed. They can describe an abstraction they built and where it leaked. Be wary of anyone who treats portability as only a rewrite, or quotes speedups without saying how they were measured.

Why it is hard to fill

Almost nobody holds this title yet. The work usually sits inside compiler, kernel, inference or infrastructure jobs, so CVs rarely say portability. The people who can do it need depth in compilers or systems and real accelerator experience on more than one platform. That mix is uncommon, and most candidates lean heavily towards a single vendor's stack. Expect to judge on evidence of work, not job titles.

Where to find them

Look at compiler engineers who have worked on intermediate representations and code generation, and at systems programmers who moved towards accelerators. Performance engineers on inference stacks are another source. They tend to sit in model providers, chip and accelerator companies, cloud infrastructure teams and open source compiler projects. Contributors to kernel languages and backend work often show their ability in public repositories.

Indicative pay

Not enough published evidence yet to give a fair range for this role, which is common for work this new. We benchmark it for you as part of a search.

Interview questions for an AI Performance Portability Engineer

  1. Describe a model you moved between hardware targets. What did you measure first, and what did you find?

    A strong answer names the profiling method, the main sources of loss and how they were confirmed. It shows a clear order of work. Worry if the answer jumps straight to rewriting kernels without evidence or baselines.

  2. How do you decide whether to write a portable kernel or a backend-specific one?

    Good answers weigh how hot the operation is, maintenance cost, and the performance gap on each target. They accept some specific code where justified. Worry about dogmatic answers in either direction, or no mention of maintenance.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We build and serve AI models, and we want them to run well on more than one kind of hardware. We are hiring an engineer to keep model quality and speed intact across the devices we support. You will work with researchers, kernel engineers and infrastructure teams.

What you will do

  • Port priority models to new hardware backends and verify quality and speed
  • Profile workloads to find where performance is lost on each target
  • Write or adapt kernels and compiler passes to close important gaps
  • Build and maintain cross-device benchmarks and regression checks
  • Design hardware abstractions that keep backend-specific code contained
  • Document which operations are portable and where limits remain
  • Advise researchers on model choices that affect portability

What you will bring

  • Shipped work moving performance-critical code between hardware targets
  • Hands-on GPU or accelerator programming, with profiling experience
  • Working knowledge of compiler intermediate representations and code generation
  • Strong grounding in computer architecture and memory behaviour
  • Experience checking numerical agreement and quality after optimisation
  • Clear written explanations of performance trade-offs

Nice to have

  • Contributions to open source compiler or kernel projects
  • Experience with distributed inference or training systems
  • Work on more than one accelerator platform

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Skills that matter

  • GPU programmingAdvancedOptimise parallel accelerator kernels
  • Hardware and computer architectureAdvancedDesign computing structures and hardware interfaces
  • Linear algebra and optimisationWorkingReason about matrix operations and objective functions
  • Distributed systemsWorkingDesign reliable work across networked machines

Other roles in aI compilers, kernels and runtime systems

Hiring nowAI Inference Compiler EngineerAn AI Inference Compiler Engineer makes trained models run faster and cheaper on real hardware. They change how model computation is compiled and…Hiring nowGPU Kernel EngineerA GPU Kernel Engineer writes and tunes the low-level code that runs model operations on accelerators. Faster, correct kernels cut compute cost and…Hiring nowML Runtime EngineerAn ML Runtime Engineer builds the scheduling, memory and execution layer that runs models on real hardware. Their work decides whether a model…

Related roles in other families

Hiring nowAI Accelerator ArchitectAI accelerators and hardware co-designAn AI Accelerator Architect decides how a chip or compute system should be built for specific training or inference workloads. Their choices set the…Hiring nowAI Hardware Verification EngineerAI accelerators and hardware co-designAn AI Hardware Verification Engineer proves that accelerator logic and its interfaces behave correctly before silicon is committed. Escaped defects…Hiring nowHardware-Software Co-Design EngineerAI accelerators and hardware co-designA Hardware-Software Co-Design Engineer shapes model software and chip interfaces together, so neither is tuned in isolation. For an AI company, that…EmergingNeuromorphic Computing ResearcherAI accelerators and hardware co-designA neuromorphic computing researcher tests whether event-driven, brain-inspired computation can beat conventional hardware on a real task. The answer…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert.

← AI compilers, kernels and runtime systemsAll new job families

Start a search

Hiring an AI Performance Portability Engineer?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.