GPU Kernel and ML Systems Researcher (Research Coding Tasks)

$80/hr

rate as listed

Workload: ~15h/week
Engagement: Project-based
Location: Worldwide
Field: STEM and Coding

Convert published ML systems and GPU kernel research into coding tasks that frontier AI agents must solve and optimize. The work includes benchmark design, hidden reference implementations, reproducible packaging and review of agent submissions.

The work

Start from a recent ML systems or GPU kernels paper with released code, ideally your own. Remove the implementation of its core contribution, then package the remaining repository as a task in which an AI agent must reconstruct that contribution and exceed the authors' results on the paper's benchmarks.

Evaluation design

You will write the task statement, create held-out workloads, set pass thresholds for each metric and provide both a hidden reference solution and a second solution demonstrating that the stated bar can be beaten. Later, you will review agent attempts for correctness, performance, robustness and compliance with the task constraints.

What afterquery screens for

Candidates need published research with released code in ML systems and GPU kernels, plus a completed or in-progress PhD in computer science or a related field. Publications at MLSys, OSDI, SOSP, PPoPP, SC or comparable venues are relevant. First-author work reporting measured improvements is preferred, as is comfort with Docker and reproducible benchmarks. Preferred technical depth includes CUDA, Triton and CUTLASS kernels, inference serving and ML compilers.

Logistics

The contract is fully remote, asynchronous and flexible. Each task typically takes a few hours, and no fixed weekly commitment is stated. Candidates should be prepared to provide publication and artifact links and to undergo the platform's background checks.

Pay band: $80/hr

What they look for

Published ML systems or GPU kernels research at a relevant peer-reviewed venue
Released code for at least one published paper
PhD in computer science or a related field, completed or in progress
Ability to design held-out benchmark workloads and per-metric pass thresholds
Ability to implement and assess reproducible GPU or ML systems experiments

Included with this listing

Interview prep — 10 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: This checks for a verifiable publication, a public artifact and direct ownership or deep familiarity with the implementation.
A strong answer: Identify one specific paper, venue, year, authorship role and repository or artifact link. Briefly distinguish the paper's core contribution from the surrounding infrastructure that could remain in the task package.
Where people lose it: Naming a research area or private repository without a paper citation, released artifact or clear explanation of the candidate's contribution loses points.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

STEM AND CODING on Sidequest

What stem and coding AI work pays

Rates by specialty, the platforms hiring, and every open role in this field.