Detecting Soft Skills in ML Engineering Roles CVs
This paper introduces an LLM-based pipeline to analyze 300 ML engineering CVs, revealing that candidates predominantly articulate soft skills through narrative rather than keywords and validating eleven hypotheses that challenge prior demand-side assumptions about role-specific competencies and seniority progression.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of technology as a massive, bustling construction site. On this site, you have three main types of builders: Software Engineers who build the sturdy scaffolding and walls, Data Scientists who design the blueprints and analyze the materials, and ML (Machine Learning) Engineers who install the smart, self-learning robots that make the building work. For a long time, people thought the only thing that mattered on this site was how well you could use your tools—your hammers, your code, your math. But recently, everyone realized that even the best builders can't finish a skyscraper if they can't talk to each other, lead a team, or figure out how to solve a problem together. These "people skills" are called soft skills.
Usually, when we try to understand these skills, we look at the "hiring managers"—the site foremen who write the job ads. They say, "We need great communicators!" But this paper asks a different question: What do the builders themselves say when they hand in their resumes? Do they actually write about their teamwork and leadership, or do they just list their tools? To find out, the researchers used a special kind of "smart robot reader" (an AI) to scan 300 real resumes. They wanted to see if the builders were hiding their soft skills in long stories about their work, or if they were just listing them like ingredients in a recipe.
The Great Resume Detective Story
The researchers, led by Aidin Azamnouri and his team, decided to play detective. They gathered 300 resumes from ML Engineers, Data Scientists, and Software Engineers. Instead of just counting how many times the word "communication" appeared, they used a powerful AI to read the stories inside the resumes. Think of it like this: if a resume is a treasure map, a simple keyword search is like looking for the word "gold." But the AI they used actually read the sentences to see if the map described how the treasure was found, even if the word "gold" wasn't written down.
They set up 13 specific guesses (hypotheses) based on what hiring managers usually say, and then they tested them against the real resumes. Here is what they found:
1. The "Storytelling" Secret
The biggest surprise was how people hide their skills. Most resumes (about 59%) didn't list soft skills in a special "Skills" section at all. Instead, candidates wove them into the stories of their past jobs. For example, instead of writing "Leadership" in a list, a candidate would write, "I led a team of five to finish the project early." The study found that candidates were three times more likely to tell a story about their skills than to just list them as keywords. This is a huge deal because many computer programs used by companies to scan resumes only look for the keywords. If you only write stories, those programs might think you have no skills at all!
2. The "Leadership" Misunderstanding
Hiring managers often say, "Everyone needs leadership skills, no matter what job they do." The researchers guessed this would be true for all three types of engineers. But the data said no.
- Data Scientists and ML Engineers talked about their leadership skills quite a bit (about 42–47% of them).
- Software Engineers, however, rarely mentioned leadership in their resumes (only about 26%).
This suggests that while software engineers might have leadership skills, they aren't writing them down as often as their peers. It's like a group of chefs where two groups brag about their knife skills, but the third group just quietly chops vegetables without saying a word.
3. The "Seniority" Shift
As builders get older and more experienced (moving from "Junior" to "Senior"), their resumes change. The study found that senior professionals are almost three times more likely to mention leadership than junior ones. But here is the cool part: they don't stop talking about teamwork. Instead, they talk about both teamwork and leadership. It's like a musician who starts by playing simple notes, but as they get famous, they start conducting the orchestra and playing the solo. The study proved that experience doesn't replace collaboration; it adds leadership on top of it.
4. The "Silent" Builders
The researchers also noticed that some resumes were completely silent on soft skills. About 18% of ML Engineers and 16% of Software Engineers had no detectable soft skills mentioned at all. However, this wasn't because they lacked the skills; it was mostly because these groups had a lot of very young, junior workers who hadn't learned yet how to write about their people skills. Data Scientists were the best at showing off their soft skills, with only 4% of their resumes being "silent."
What This Means for Everyone
The paper concludes that there is a "gap" between what candidates do and what they write. The skills are there, hidden inside the stories of their work, but the old-school computer systems that scan resumes are too dumb to find them. They are looking for a list of ingredients, but the cooks are writing a novel about the meal.
The study suggests that if you are a young engineer trying to get a job, you need to stop just listing your tools. You need to tell the story of how you used those tools to help people, lead teams, and solve problems. And if you are a company using robots to scan resumes, you might be missing out on the best candidates because your robot is only looking for keywords and ignoring the stories.
In short, the paper didn't just count words; it proved that the way we talk about our soft skills is changing, and if we don't update how we read them, we might be missing the best builders on the site.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.