Bringing Silicon Valley to Kathmandu Valley

For those who think that Nepal is too underdeveloped to make full use of artificial intelligence (AI), think again. That is exactly what they used to say about computers and mobile phones in the 1990s.

It may come as a surprise to many that Nepal has been gaining ground in AI, developing not only software using machine learning algorithms but producing world-class engineers. One company at the forefront is Fusemachines Nepal, which has started using industry experts to train AI students with cutting-edge technology to deliver intelligent solutions.

“I wanted to see if I can contribute in bringing the best AI education to Nepal and make Nepal known around the world as one of the best sources of AI talent,” says the Nepali founder of Fusemachines, Sameer Maskey, a professor at Columbia University.

This is the age of surveillance capitalism, where algorithms determine election outcomes, Siri knows what you want before you do, wearables correctly deduce the state of the heart and Facebook recognises friends.

Read also:

Natural Intelligence, Duksangh Sherpa

Kathmandu's Silicon Alley and the Law, Sonia Awale

AI simply imitates human thinking by recognising patterns in data, so that repetitive everyday work can be done by machines that learn as they go along.

Nepal missed the bus on natural resource processing, manufacturing and information technology. But experts say that training a critical mass of engineers in AI can allow the country’s economy to leapfrog and become globally competitive.

Fusemachines Director of Academic Affairs Bülent Uyaniker, who was in Nepal recently, rejects the notion that Nepal is not ready for artificial intelligence applications. “It is happening already, it is inevitable. If there can be 8.5 million Facebook users in Nepal, then it has the special conditions for AI.”

Proof of this is the increasing number of software companies in Nepal using local engineering talent to work on software solutions for customers in North America or Europe. However, most of the engineers and recent graduates need training in AI to keep up with customer requirements. America alone will need 200,000 data scientists in the next five years, and most of these will come from the UK, Finland, Canada, Singapore, China and India.

Which is why Fusemachines Nepal is also emphasising education. Says the head of its global operations and strategy, Sumana Shrestha: “You cannot learn AI in a one-day bootcamp, it needs intelligent mathematics, but there is a huge demand versus supply gap for engineers proficient in machine learning or other AI components everywhere.”

Read also:

It's all about IT, Sahina Shrestha

Nepal established itself as a sought after destination in the past 20 years for outsourcing services such as software and app development, website design and big data management to overseas clients, mostly due to the country’s inexpensive English-speaking workforce.

This move from IT to AI will not just create jobs in Nepal, but also allow the country to increase efficiency and productivity in the workplace. General practitioners in rural hospitals will be able to make diagnoses faster so they can spend more time with patients, high-risk individuals can be identified with cancer screening, and targeted advertising and customised itineraries will lure potential tourists during Visit Nepal 2020.

Recently, a group of engineering students developed a model to help poultry entrepreneurs understand fowl behaviour and the state of their animals’ health, helping them to raise the farm’s business profile.

“With precision livestock farming we can generate patterns to help farmers recognise symptoms before an outbreak of a disease by implementing AI components such as image processing and deep learning,” explained engineering student Sajil Awale at Pulchok Engineering Campus. “This allows for timely intervention to prevent mass deaths and reduce losses.”

Read also: Tap and tour, Clara Bullock

Computer vision (which enables computers to see and process images as humans would) can also help identify rotten fruit swiftly, and prevent misuse of pesticides by identifying areas on the farm that require chemicals, and the amounts needed. AI can also estimate future harvests, allowing farmers time to find markets for produce.

Engineers at Fusemachines Nepal are working on Nepal’s first optical character recognition (OCR) system so forms filled out with Nepali handwriting can be digitised and translated into English. This will have huge scope in Nepal’s banking, hospital and government sectors, where pen and paper continues to be the norm.

Sixit Bhatta, CEO of ride-sharing startup Tootle, says Nepal is ripe for AI applications: “Our efforts now should be on preparing for a world in which machines perform skills-oriented tasks and for humans to take on the roles that require creativity and empathy. But before that, the government should design policies that allow AI to grow, and not restrict it.”

Sumana Shrestha at Fusemachines Nepal says that as long as salaries for clerical staff are low, there is less potential for AI to flourish. But she adds: “The curse of cheap labour means companies will prefer to employ people to do repetitive work. But sooner or later, AI will be here. Nepal needs to develop despite government. And the private sector needs to prepare itself for disruption.”

Coming to terms with AI

 Artificial Intelligence: Ability of computer systems or machines to make a decision like humans, or the ability to perform tasks requiring human intelligence

Machine Learning: A subset of artificial intelligence that provides a system with the ability to automatically learn and improve from experience without being explicitly programmed, relying on patterns generated from data

Deep Learning: Machine learning that is applied on a large set of data, also known as deep neural learning that uses deep neural networks to model complicated data

Natural Language Processing: Interaction between computers and human languages, deals with programming computers to process and analyse natural (human) language, this field of AI processes, analyses, interprets and distills information from human languages

Computer Vision: Enables computers to see, identify and process images in the same way that human vision does

Image Processing: Part of computer vision that entails analysis and manipulation to find insights from a digitised image

Big Data: Extremely large data sets on which AI is applied to reveal patterns, trends and associations and make decisions

Making Nepalis artificially intelligent

During a visit to Nepal six years ago, Columbia University professor Sameer Maskey handpicked three students from engineering schools across Kathmandu who were able to solve a mathematical equation. The three went on to become the core of Fusemachines, a global company that aims to democratise artificial intelligence (AI) through education and software solutions.

Headquartered in New York, Fusemachines has since opened branches in Canada, UK and the Dominican Republic to develop intelligent software solutions that have transformed brands and businesses around the world. One of its biggest operations is in Nepal, Fusemachines employs 100 top Nepali software engineers who work on projects that use AI applications in fields ranging from telecommunications and banking to hospitals and governance.

Unlike other back office companies that work on outsourced software development, Fusemachines is a school in itself, training engineers while coming up with product solutions.

“We employ senior engineers and industry experts with PhDs along with upcoming engineers, who work together to solve client-specific problems though AI,” explains Sumana Shrestha who heads Fusemachines’ global operations and strategy. “Such collaborative approach allows young talent to continuously learn and grow.”

Following Maskey’s vision, Fusemachines tries to make AI accessible to everyone through education, which is why it initially offered training fellowships and then, to meet the demand for engineers, launched AI Shikshya — a year-long, in-house training program.

“With our own proprietary platform and content we have partnered with engineering colleges in Kathmandu to offer AI,” says Shrestha. “The program is a blend of online and on-site, the course material is not too academic, is industry focused and instructors are seasoned engineers up to date with new algorithms.”

The company also has its own AI schools, open to professionals who want to grow their business or anyone interested. Fusemachines also offers a foundation course, open to high school graduates. A year-long, micro-degree program consisting of four major courses — including machine learning, deep thinking, natural language processing and computer vision — will be the next step.

“In school we were always presented with clean data sets to work on but that is rarely the case on the job. But this year-long training program gets people ready to take on real-life problems and provide AI solutions,” says Rojesh Shikharkar, engineer at Fusemachines Nepal and a post-grad at Pulchok Engineering Campus.

Apart from training, Fusemachines has given engineers who would otherwise have migrated to work in the US and Europe an opportunity to find meaningful work in Nepal.

Says Shrestha: “If you create opportunities here, people might actually stay, and once we have mass education in artificial intelligence, companies here will start adopting and see the value of AI products, ensuring more opportunities at home.”