Runtime and Performance Engineer

Desert Ant Labs; Amsterdam; Hybrid; EUR 90,000 - 140,000; Full time

Benefits offered by this job

Office in Amsterdam

Remote work option

Job summary

Desert Ant Labs seeks a developer to optimize models for Apple platforms, Android, and Windows, shaping the Swift core and platform bindings. You will push workloads to GPUs and Neural Engines, and design SDK APIs to maximize speed on diverse devices.

You will write Metal/WebGPU kernels when needed and collaborate with researchers to advance model architectures for real devices, contributing to a fast, robust SDK.

Qualifications

• Experience speeding software on real hardware and profiling performance.
• Shipped an SDK or runtime used by developers.
• Able to read compiler output and profiler traces.
• Write Swift with C interop and bindings to other languages.
• Work in C, C++, Kotlin, or Rust where a platform needs it.
• Knowledge of GPUs/ML accelerators and kernel programming.
• Contributed to llama.cpp, MLX, or ONNX Runtime.

Responsibilities

• Run as much of each model on the Neural Engine, NPU, or GPU as the hardware allows.
• Build the shared Swift core and its bindings to Kotlin, WebAssembly, and native code: memory budgets, routing, and fallbacks.
• Write Metal, WebGPU, or native kernels when a runtime runs slowly; you may write the first ones.
• Collaborate with researchers on model architectures that run fast on real devices.

Skills

Swift

Kotlin

C++

Rust

WebAssembly

Tools

Metal

DirectX

WebGPU

ONNX Runtime

llama.cpp

MLX

Job description

Make our models run faster on Apple platforms, Android, and Windows, in our shared Swift core and its platform bindings.

• You have deep experience on Apple platforms (Swift, Metal), Android (Kotlin, the NDK), or Windows (C++, DirectX).

Location Amsterdam, Remote

Time zone UTC-5 to UTC+1

Type Full time

About the role

You work from GPU kernels up to the SDK API. You also design the SDK API, so the default call runs a model the fastest way the device allows.

What you will do

• Run as much of each model on the Neural Engine, NPU, or GPU as the hardware allows.
• Build the shared Swift core and its bindings to Kotlin, WebAssembly, and native code: memory budgets, routing, and fallbacks.
• Write Metal, WebGPU, or native kernels when a runtime runs a model too slowly. We have none yet, so you write the first.
• Work with researchers on model architectures that run fast on real devices.

You might be a fit if you

• Have made software measurably faster on real hardware, and can explain the profile and the fix.
• Have shipped an SDK or runtime that other developers built on, and have changed an API after seeing developers use the API wrong.
• Read compiler output and profiler traces.
• Write Swift, including C interop and bindings to other languages.
• Work in C, C++, Kotlin, or Rust where a platform needs it.
• Write clean, tested code, and spot a weak change in review, whether a person or an agent wrote it.
• Plan and run your work through coding agents such as Claude Code, with a low tolerance for slop.
• Contributed to llama.cpp, MLX, or ONNX Runtime.
• Wrote GPU compute code on mobile.

How we work

Start what needs starting without waiting to be asked, and finish what you start. Take on work outside your role when a project needs you.

We ship quickly, so we cut scope until only the part users notice is left. Anyone can comment on your work or redo your draft, and we say early when work isn't ready. We read that feedback as help.

We judge the work by what shipped and what changed because of it. Nobody counts hours.

What we offer

• In our Amsterdam office, or remote anywhere between Eastern Time in North America (UTC-5) and Central European Time (UTC+1), so everyone shares part of the working day.