The PhD Trap
Summer hiring is in full swing, and job boards are flooded with AI roles. Most of them are wrong. Companies that have never trained a model are hiring ML researchers. Startups that need to ship product are hiring people who want to write papers. The mismatch between what organizations need and what they are recruiting for has never been wider.
Here is the uncomfortable truth: most companies do not need someone who can build a model from scratch. They need someone who can build products that use models effectively.
The Three AI Roles That Actually Matter
- AI Product Engineer. This person understands model capabilities and limitations, can design prompts and pipelines, handles evaluation and testing, and ships production-grade AI features. They are part software engineer, part product thinker. This is the role 80% of companies actually need first.
- AI Strategist. This person can evaluate which problems are worth solving with AI, estimate build vs. buy tradeoffs, and translate business needs into technical requirements. They prevent the organization from building impressive things that do not matter.
- ML Engineer. This person fine-tunes models, builds training pipelines, and optimizes inference. You need this person only after you have validated that custom models outperform off-the-shelf options for your specific use case. For most companies, that is not day one.
Why This Mistake Keeps Happening
Three forces conspire to create the wrong hire. First, the AI talent market is confusing. Titles are inconsistent and credentials are noisy signals. A Stanford PhD who spent five years on theoretical NLP might be less effective at shipping AI products than a senior engineer who has been building with Claude and GPT for two years.
Second, hiring managers default to prestige. They hire the most impressive resume instead of the most relevant skillset. Third, companies conflate AI ambition with AI readiness. They hire for the team they want in two years, not the team they need today.
The best first AI hire is not the smartest person in the room. It is the person who can ship something valuable in 90 days.
How to Get It Right
Before you write the job description, answer one question: what is the first AI-powered capability you want in production, and what skills does it actually require? Work backwards from the deliverable, not forward from the org chart. The answer will almost certainly point you toward an AI product engineer, not a researcher.