How To Vet a Prompt Engineer

Skilled technologists have been working in and advancing the field of machine learning since the 1950s—but recent consumer access to generative artificial intelligence (AI) has changed the playing field forever.

Leveraging machine learning (ML) frameworks is no longer something that only the top tech companies can do. With tools like OpenAI's ChatGPT, anyone can leverage AI for their business and unlock new levels of productivity.

If they know how, that is.

These tools require careful prompting, in the form of text-based inputs, to deliver high-quality outputs. And the learning curve is steep.

As a result, many companies realize that rather than reskilling their entire workforce—which will take a long time—they can quickly reap the benefits of GenAI by working with an expert in prompt engineering.

But finding a prompt engineer can be tricky. Prompt engineering is such a new field that you won’t find university degree programs that have been fine-tuned over decades. And it's hard to assess someone’s hard skills in a brand-new field that you yourself aren’t experienced in.

You’re going to want to work with prompt engineers as soon as possible, though, in order to maintain a competitive edge—so it’s imperative to figure out what, exactly, makes a prompt engineer great.

At Upwork, we’ve been working on vetting prompt engineers as part of our Expert-Vetted Talent program. And today, I’m sharing some of our learnings and takeaways that can help you better vet prompt engineers for your own projects.

Identify why you need a prompt engineer

First, it’s important to have a clear idea of why you want to work with a prompt engineer. This will point you in the right direction as you begin your search for a qualified professional.

Think about it a bit like this: If you need to give instructions to your five-year-old, you won’t use the same words and phrases that you do when giving instructions to a colleague or a peer. There’s a certain vocabulary you know you need to use for your child to understand you.

In the same vein, you can’t communicate with a generative AI model the same way you’d talk to your direct reports at work. While these tools can interpret human speech through natural language processing (NLP), certain words and patterns will help guide the AI toward the desired response.

A prompt engineer understands these nuances and can “talk” to generative AI systems in just the right way to get results. They know when they need to give the AI examples, how to provide feedback, and even how to adjust its level of creativity—also known as its “temperature.”

With the help of a prompt engineer, you could more easily:

• Fine-tune a generative AI tool to help you create social media captions that capture your brand voice
• Use AI to direct your customers to online help resources when they contact your customer service team after hours.
• Summarize in-depth research reports and pull out statistics that will appeal to your audience.

Types of AI prompt engineers

Prompt engineering isn’t a catchall, either. The field has several sub-specialities. Pros in these application areas may focus on leveraging generative AI to:

• Speed up software development
• Process large amounts of data
• Create written content
• Produce artwork or graphic design concepts

You may even find that you need more than one prompt engineer for some projects.

If you’re trying to use generative AI to speed up the process at which your team can produce Instagram content, for example, you may need one prompt engineer to help you leverage design tools like Midjourney—and another to help coax witty captions out of Anthropic’s Claude.

Key skills for prompt engineers

Once you establish your goals for working with generative AI, you can begin assessing prompt engineers to find the right fit.

You’ll need to look at their hard skills, specific soft skills, and overall experience.

Hard skills

When you’re looking for a prompt engineer, you’ll want to find someone who has technical skills in both prompting and using machine learning models.

Look for prompt engineers who have experience with:

• Large language models (LLMs): LLMs are the data-rich engines that allow generative AI tools like ChatGPT to guess what you want it to produce based on your inputs.
  
  
• NLP principles: Generative AI tools parse sentences—much like how you might diagram one—but look for different parts of speech than humans do. A great prompt engineer has a deep understanding of how AI assesses sentiment, identifies named entities, and assembles structured information into plain language.
  
  
• Python: Python is the primary development language used in AI and ML applications. Prompt engineers who aren’t focused on code generation may not be “fluent” in Python, but should understand key basics.
  
  
• Machine learning frameworks: A variety of different ML frameworks—including Apache MXNet, TensorFlow, and PyTorch—are used to train AI tools. An understanding of how specific frameworks operate can be useful in crafting the perfect prompt.
  
  
• Machine learning techniques: Prompt engineers may use a few different techniques to train an AI on what they want a tool to do—listen for terms like “few-shot learning” and “tree of thoughts'' or “chain of thoughts.” These are a few first-level approaches to prompting a generative AI tool to deliver results.
  
  
• Your generative AI tool of choice: Different AI tools require different prompts. The way you prompt a tool for AI image generation can vary between platforms, and ChatGPT might require different prompts from a tool like Jasper.

And, of course, you’ll also want to make sure that the prompt engineer you select has a skill set related to how you plan to use your prompts.

Some prompt engineers have previously worked as writers, developers, data scientists, artists, and designers. Others have always worked in machine learning, but specialize in coaxing a specific result—code, art, data or words—out of AI, so they understand what a quality result looks like.

For example, if you need a prompt engineer to help you with using an AI solution to write code, you’ll look for someone who:

• Has worked as a software engineer
• Has professional experience or education in machine learning and AI technologies
• Understands Python along with the languages you and your team plan to generate code in
• Demonstrates recent experience engineering effective prompts for projects related to coding and development

Soft skills

Every prompt engineer also needs soft skills—ones that come with the territory of working in a new and quickly evolving field.

A successful prompt engineer will show strong:

• Problem-solving skills: They aren’t just reacting to something; they’re actively working to develop solutions
• Tenacity: They’re able to keep testing, trying, and iterating until they find the right solution
• Flexibility: They’re able to pivot and flex when things don’t go as originally planned
• Curiosity: They’re committed to continual learning as ML and AI continue to develop
• Project management: They’re able to plan the stages of a project that need to happen in order to reach the end goal, and can explain how they’ll get from point A to B
• Customer service: They’re able to guide clients through working with AI, regardless of how much prior experience the client has with such systems

A prompt engineer will also have strong communication skills—both with you, their human collaborator, and with the AI.

Experience

Finally, as with hiring for any role, you’re going to want to make sure the prompt engineer you work with has relevant experience.

You don’t need to be an AI pro yourself to evaluate this, either. Look for:

• A client list that shows they’ve worked with companies of a similar size and scope to yours
• A portfolio of work or case studies that show off the results they’ve been able to deliver for their clients
• Good client reviews or testimonials

If you’re hiring a prompt engineer on Upwork, you can see details about their experience—and prior client feedback—right on the talent’s profile page.

And if you’re not hiring through Upwork, or simply want to do more research on a prompt engineer, checking out their LinkedIn or GitHub presence is a good place to start.

‍

What isn’t necessary

Formal education isn’t a requirement for becoming a prompt engineer—but several specific fields of study can help. Machine learning, computer science, data science, and communications are all great backgrounds for a prompt engineer to have.

Don’t worry too much about the level of degree they’ve achieved, though. Their hands-on experience is really what’s key here.

Generative AI has grown so quickly that someone who earned a PhD in machine learning in 2015 might be less skilled at prompt engineering than a recent computer science graduate who’s been immersed in generative AI on the side for the past year.

In terms of education and experience, it’s most important that a prompt engineer is:

• Keeping up to date with changes and developments in the world of AI by reading documentation and attending conferences
• Researching and testing out new developments on their own
• Pursuing ongoing education through accredited, tech-focused schools
• Actively working with new AI tools that hit the market

That said, you can definitely tap high-level machine learning experts—like those PhDs—for help understanding how to vet a prompt engineer, even if they aren’t offering prompt engineering services themselves.

Someone with a background in machine learning can help you figure out exactly what you’re looking for, and the right questions to ask.

Questions to ask a prompt engineer

Ultimately, you’ll need to tailor your prompt engineering interviews to reflect your company’s tools and goals—but these 16 questions can serve as a great jumping-off point.

1. Can you explain a bit about how NLP and LLMs tie into the prompt development process?
2. What entails a good prompt? What makes a prompt bad?
3. How do you evaluate whether a prompt works or not?
4. What programming languages do you work in?
5. What are your preferred machine learning frameworks?
6. How experienced are you with the AI application we’d like to use?
7. What machine learning techniques do you use when crafting a prompt?
8. What’s your process for making sure a prompt is clear, actionable, and guides the AI to produce the desired result?
9. Can you explain how you’ve successfully developed prompts for clients in the past?
10. How would you approach structuring a prompt that would help someone use AI to evaluate the sentiment in customers’ product reviews?
11. Tell us about a time when your prompts didn't work as expected. What did you do to fine-tune them?
12. What's your approach to ethics when crafting prompts and training machine learning models?
13. How do you balance creativity and accuracy in the AI’s output?
14. How are you staying up to date with new advancements in AI?
15. How can you help us figure out the right kinds of prompts for our goals and use case?
16. What kinds of support are you able to give our team as we learn to implement your custom prompts?

Where to find skilled prompt engineers

You can hire prompt engineers on Upwork right now—our skill labels and filters make it easy to find pros with the right prompting experience for your project.

Simply post a prompt engineering job or begin reaching out to professionals who interest you. If you’re an Upwork Enterprise customer, you can even skip part of the vetting process by browsing the profiles of Expert Vetted Talent experienced in prompt engineering—or by working with your Upwork Professional.

Our team has done the initial vetting for you by making sure that these pros are truly experts in the field of prompt engineering. All you’ll need to do is make sure they’re the right fit for your project and budget. Log into your Enterprise account now or reach out to our Upwork Enterprise team to get started.

‍

Upwork does not control, operate, or sponsor the tools or services discussed in this article, which are only provided as potential options. Each reader and company should take the time to adequately analyze and determine the tools or services that would best fit their specific needs and situation.