It’s unusual for businesses to interact with students in a major university class, and its an important reason why so few students graduate without having solved real world problems. But, Johns Hopkins University and online course provider Coursera, may have found a novel solution to bridge business and students.

With an eye on training the next generation of data gurus, Johns Hopkins has been running a free data science course through Coursera and just announced a partnership with tech startup Swiftkey, which will design and tutor students as they complete their final homework assignment (a “capstone”).

“This is a completely new way to scale the education of data scientists,” professor Jeff Leek writes to me. For a few months, I’ve actually been enjoying Leek’s class in my ongoing review of data science courses across the big online course providers. In some ways, it’s comparable in quality to the statistics graduate courses I took at the University of California at Irvine.

But massively open online courses (aka MOOCs) don’t scale hands-on learning well. This is where Coursera thinks it can find a profitable fit between online education and vocational learning. Through its “signature track” service, Coursera charges students for access to the capstone project and a verified certificate of completion. Leek’s courses are $50 a course, or $500 for the whole data science track.

Up until now, the capstones’ pretty much resembled the kinds of homework I encountered in grad school: find a dataset, run some code used in class, and write a report. Swiftkey, a London-based startup that designs auto-completion software for keyboards, will help design a final homework assignment that mirrors the data problems students will encounter in the real world.

“We are going to give our learners unstructured data and a vague set of instructions, just like they would encounter in industry, and let them demonstrate the skills they have learned along with their personal creativity,” Leeks wrote. The projects are available this fall.

At an actual university, it’d be unusual for a statistics professor to let a startup design homework problems. Academics generally thumb their noses at vocational training. Coursera, it seems, has found a way to let in some real-world skills.

For anyone thinking about taking the data science series, note that some of the courses may be difficult for beginners. I tried my hand at the data-wrangling course and found myself in the middle of a tiny revolt of students who found it too difficult. I haven’t found any courses online yet that I think can take a beginner completely through the process of becoming a data scientist without a lot of frustrating self-teaching outside of the curriculum.

That said, I still found it valuable to take the courses and do recommend them for both beginners and those who want to refresh their skills (I also enjoy Udacity’s data science series so far).

If the partnership proves successful enough, it might help Coursera pay back the massive rounds of funding it has received and find a path to profitable education for all.