Software/ML Engineer

What you'll do

AfterQuery is assembling senior engineers to test how well frontier models reason about LLM inference systems, GPU-level performance optimisation and model-serving architecture. You design realistic scenarios and problem sets, write reference solutions, and grade model output on real systems-engineering problems.

The ad is explicit that this is a research-and-evaluation role, not a traditional engineering job: you will not ship production code for AfterQuery; you define what correct and excellent look like for models tackling problems you already know.

Who fits

• A Bachelor's degree or higher in computer science, computer or electrical engineering or a related field
• Hands-on background in one or more of GPU inference optimisation, custom CUDA kernels, model-serving systems (vLLM, TensorRT-LLM, SGLang), quantisation or optimiser design
• Experience at a recognised technology or AI company, in a role with real systems ownership
• Strong written communication, since you author explanations and feedback

Preferred: direct experience with SGLang, Mamba/Mamba2 or IBM Granite-family models; open-source contributions to inference or serving frameworks; distributed inference experience (tensor and expert parallelism, continuous batching, KV-cache optimisation).

What it pays

The board shows $125/hour. The ad describes the role as remote, contract, part-time and flexible, and gives no weekly hours or payment terms; the AfterQuery company page covers how payouts and assessments work in general.

Worth knowing

Good:

• High rate for a part-time, self-scheduled role
• Very specific, in-demand subject matter
• Evaluation work rather than production delivery

Less good:

• The experience bar is steep and narrow
• The ad does not state hours, volume or duration
• Assessments and onboarding are not described in the ad

About this listing

Read from the AfterQuery Experts board on 18 October 2026; posted 22 July 2026. See AfterQuery.