How to Write Job Descriptions That AI Engineers Actually Read

<p>The average job description is written for HR compliance, not for the candidate. For AI engineers — who receive 5-10 recruiter messages per week — a bad JD is an instant pass.</p> <h2>Lead with the Problem, Not the Company</h2> <p>Engineers care about what they'll build. "We're looking for a Senior ML Engineer to improve our recommendation system's latency by 10x while maintaining accuracy" is infinitely more compelling than "We're a fast-growing startup looking for talented engineers."</p> <h2>Be Specific About Tech Stack</h2> <p>Vague requirements like "experience with ML frameworks" tell candidates nothing. Specific requirements like "Production experience with PyTorch, familiar with distributed training on multi-GPU setups, comfortable with ONNX/TensorRT for inference optimization" attract the right people and self-screen the wrong ones.</p> <h2>Include Compensation</h2> <p>In 2026, salary transparency isn't optional — it's legally required in many jurisdictions and expected everywhere. Companies that include compensation ranges get 3x more qualified applicants.</p> <h2>Kill the Requirements Laundry List</h2> <p>When you list 15 requirements, you get zero applicants who meet all of them and lose the ones who meet 12. Separate into "must have" (3-5 items) and "nice to have" (the rest).</p> <h2>Show, Don't Tell, About Culture</h2> <p>"We have a great culture" means nothing. "Engineers ship to production on day one, we do async code reviews within 4 hours, and we have a monthly demo day where anyone can present" shows what working there actually looks like.</p>