When you don’t know how to do a picture, add a cat
Machine learning techniques are becoming a common part of every software engineer’s toolkit.
At the beginning of any new field, there will be special jobs, and as the field grows, those special jobs will become more common. The field of machine learning may be following this pattern.
Let’s break it down
The rise of machine learning engineers is the result of corporate buzzwords like ARTIFICIAL intelligence and data science. Machine learning engineers were a necessary role in the early days of machine learning, and for many of us, the job provided a nice increase in income! But if you ask machine learning engineers exactly what they do, you’ll get different answers from different people.
Purists would say that the job of a machine learning engineer is to “rescue” models from the lab and put them into production for practical use. Machine learning engineers scale machine learning systems, turn reference implementations into usable software, and often incorporate data engineering content into them. Machine learning engineers are often powerful programmers with some understanding of the models being used.
But these descriptions sound a lot like what your average software engineer would do.
If you ask top tech companies what machine learning engineers mean to them, you’ll probably get 10 different answers from 10 survey participants. The results aren’t surprising, machine learning engineer is a relatively young position, and the people Posting these positions are often managers who may not have had the time (or inclination) to really understand the technology in the tech world for decades.
I’ve compiled a list of requirements for machine learning engineers posted by several tech companies to see how they differ from one another:
Ph.D. in mathematics/Statistics/Operations Research. Master of R, SQL and machine learning skills. The first JD is interesting. Is it a machine learning engineer, not a researcher?
Bachelor’s/Master’s degree in Computer Science. One to five years of working/academic experience in software development, exposure to computer vision, natural language processing, etc. is a plus.
It’s also typical to insist on “learning background”, but this JD comes from a head tech company, so it’s not surprising.
Bachelor’s/Master’s degree in Computer Science. Three years or more experience in programming and building machine semester systems. Familiarity with big data is a plus.
This last one is more in line with what we think of as a machine learning engineer.
Some companies have already started using a new approach to Posting job requirements, and I think most will follow suit. At the heart of the approach is the requirement that applicants be software engineers familiar with machine learning techniques, preferably with several years of work experience. Employers will prefer engineers with extensive experience in building and scaling systems, whether based on machine learning or other technologies.
As long as little is known about machine learning and the barriers to entry are high, the job of machine learning engineer will be necessary.
After serious consideration, I believe that the role of the machine learning engineer will be completely replaced by that of the common software engineer, and that the machine learning engineer will function like a standard engineer, taking requirements or reference implementations from upstream people, turning them into production code, delivering and extending them into applications.
At the moment, many of us machine learning engineers are in the strange position of using machine learning techniques to solve problems they have never encountered before. So, in a lot of cases, they’re doing research and engineering on the side. I know people in the industry who are good at both. Others are less skilled, but spend more time reading new research papers and exporting what they’ve learned into usable code
Faced with an awkward intersection, we as machine learning engineers are also trying to find the right location for us.
Influenced by the way they work, machine learning engineers often participate in discussions and conferences, and we accept invitations to conferences whether or not the topics are core to our expertise… It seems to me that machine learning engineers are usually at the very end of building reference implementations, and are then solely responsible for converting reference implementations into production code.
Before long, most enterprises will have little need to leverage research efforts to improve their projects, with special skills being used only in rare cases and for deep technical work, and engineers making extensive use of apis. As more universities bring machine learning knowledge into the classroom, machine learning is becoming a common tool in every new engineer’s toolkit. University courses on machine learning are almost full, so almost every graduate knows something about the field.
Here are some of my thoughts on the questions raised by many column readers who leave comments behind the scenes: Maybe the “universal API” of Silicon Valley is a lie, or maybe artificial intelligence will always be needed to make custom adjustments when building system infrastructure. But I think most of the problems in other areas will be solved by a simple API, just as HuggingFace did for natural language processing. That’s my personal opinion.
Some people say, “Dude, it’s just a title, machine learning engineer is someone who has a stronger background in math and statistics than the average computer science graduate.” I totally agree. It’s just a title. But if the job itself does not exist, will the title survive? But one thing is true, it’s just a title.
I think the future development of blockchain distributed system engineers, which is hot right now, will be similar to that of machine learning engineers. Since Satoshi Nakamoto issued the Bitcoin white paper, the vast majority of blockchain projects have been committed to the construction of basic technology and infrastructure. This requires strong engineering skills from project participants, often referred to as distributed systems engineers. What you end up seeing is that this is becoming more and more abstract, as companies start to look for practical use cases, and ordinary engineers start to build new ones using blockchain technology, just as artificial intelligence and machine learning are changing.
One of my favourite messages since I posted this came from Varii on Twitter: “As you say, machine learning engineers are just a headline. Most employers want a mix of skills. “I think it’s not about who’s going to die out, but who’s going to be strong enough to adapt to the changing industry.”
Yes, this is consistent with my idea that if you are truly passionate about something, no matter what the field or trend is, your passion will drive you to create a lot of great stuff.
The original link: towardsdatascience.com/machine-lea…
The above information comes from the network, edited by the “Jd Zhilian cloud developer” public number, does not represent the position of JD Zhilian cloud
Welcome to click “read” to learn more about jingdong Zhilian Cloud!
