In the summer of 2011, I was giving a lecture on machine translation at a small college in Kathmandu. That afternoon, one student asked me a question that was so complex and required such fundamental understanding of computer science that it made me realize how some students in the most remote corner of the world — with hardly any access to advanced technology and academic privileges — have the level of talent and curiosity as some of my students at Columbia University. Two years later, I hired the student to work for me at Fusemachines, where he currently leads a team of 90 engineers.

Chinese tech giant Tencent, in a study compiled by its research institute, estimates there are around 300,000 AI professionals in the world — but millions more are needed. As artificial intelligence permeates through every field in every industry, there is a war for AI talent. In fact, Silicon Valley giants are fighting and paying an exorbitant amount of money to lure the best AI engineers to work for them. “Salaries are spiraling so fast that some joke the tech industry needs a National Football League-style salary cap on AI specialists,” Cade Metz wrote in a New York Times article.

But instead of upping the salaries to millions of dollars and fighting for the same small pool of talent, we should be training engineers in artificial intelligence around the world. Young students and engineers in remote developing countries also have the ability to perform — and, at times, outperform — the ones who have degrees from elite institutions in the West. There is untapped talent in these places, and we are neglecting it to our detriment.

Educating engineers from across the globe in machine learning, deep learning and natural language processing — the most common sub-disciplines within AI — will help increase access to AI talent. Someone who experiences complex problems in his/her own country could be more suited to try and solve those problems with AI. For example, a Nepali engineer who wants to use machine learning to predict crop yields of their community will be better informed about Nepal’s farmlands than a graduate from Silicon Valley.

Similarly, engineers working for companies like Zipline and Fusemachines in Nepal and Rwanda are able to build and adapt autonomous drones to deliver medicine in remote villages in countries with poor road infrastructure. This would not only save lives but also remarkably change the livelihood of villagers who would otherwise have to trek for days to get to the nearest pharmacy. This is one example of the numerous ways artificial intelligence can be used to improve health care, fight poverty and raise the standards of living in developing countries. For that, it’s important that we invest in educating and enabling talented young engineers in such countries.

But how do we train local engineers in far-flung places to build drones, robots and complex systems? The answer is in a combination of online courses and some on-site training. Two years ago, Fusemachines launched a fellowship program that allows students in Nepal to develop high-level skills in programming and solving machine learning algorithms — eventually leading to a MicroMasters in Artificial Intelligence from Columbia University. Today, the program has expanded to three additional locations: the Dominican Republic, New York City and Rwanda.

As they complete the course, enrolled students come to class once a week and discuss the homework assignments and problems. What we’ve found is that this mix of an online course with on-site guidance works very well with the students. They learn on their own time throughout the week but still feel like a part of a class when they meet with other students in a physical location. With this model of learning, we have had many engineers graduate with certificates in AI from Columbia University.