There is a gold rush for AI talent. But most companies are mining for the wrong resource. They put up job descriptions for "PhD in Machine Learning" or "Senior Data Scientist," hoping to find someone who can build the next GPT-5. But 99% of companies don't need to build models; they need to use them.
Enter the AI Engineer.
The Profile: The "Product" Engineer
The AI Engineer is closer to a Full-Stack Developer than a Researcher. They live at the application layer. They don't know how to do Backpropagation calculus by hand, but they know exactly how to manage context windows, prevent hallucinations, and chain prompts to get a reliable output.
They are "AI Whisperers." They understand that LLMs are non-deterministic (stochastic) engines. This requires a different mindset. In traditional code, 2+2 always equals 4. In AI code, 2+2 usually equals 4, but sometimes it equals "The color blue." The AI Engineer builds the guardrails to handle that uncertainty.
The New Stack
If you are hiring this person, check their GitHub for these keywords:
- Vector Databases: Pinecone, Weaviate, pgvector. (For Long-Term Memory).
- Orchestration Frameworks: LangChain, LlamaIndex. (For chaining steps together).
- Eval Frameworks: "How do we know the bot is good?" Tools that run automated tests on the AI's answers.
- Cloud Functions (Serverless): Because AI is event-driven.
The Interview Test
Don't ask them to invert a binary tree on a whiteboard. That is useless.
Give them a practical task: "Here is a PDF of our Employee Handbook. Build a simple RAG (Retrieval Augmented Generation) bot that allows me to ask questions about our Vacation Policy. You have 2 hours."
The good ones will not only build the bot; they will break the PDF into logical chunks, use a good embedding model, and write a system prompt that says: "You are a helpful HR assistant. If you don't know the answer, say 'I don't know'—do not make it up."
Conclusion
The companies that win will be the ones that recognize this is a new discipline. It is "Software Engineering 2.0." Respect the nuance, hire the builders, and let them experiment.