Zindi winners Darius and Nikhil help AirQo improve air quality predictions in Africa
The 2nd-place winners of Zindi’s recent AirQo Ugandan Air Quality Forecast Challenge took their data science journey to the next level recently by joining the project team to implement their machine learning solutions to improve air quality monitoring inKampala, Uganda.
“Machine learning competitions give you the rush to try out new ideas, racing against time and the leaderboard,” Nikhil Kumar Mishra says with a smile. “I am passionate about everything data science.”
His teammate, Darius Moruri, cannot hide the excitement in his voice when talking about AI. As a self-taught data scientist from Kenya, Darius says he is passionate about competing in data science hackathons.
Nikhil and Darius were brought together by their passion for data science. Not only did they team up — thousands of miles away from each other — in the AirQo Ugandan Air Quality Forecast Challenge competition, they came 2nd among more than 700 data science competitors from countries like Uganda, Tanzania, India, Nigeria, Japan, and the UAE.
Machine learning model for air quality
The competition was organised in partnership with the Digital Air Quality East Africa (DAQ EA) project, University of Birmingham and the AirQo project from Makerere University, Kampala, and focused on creating a machine learning model that would accurately predict air quality in Uganda. This competition was not just about winning the $5,000 prize fund; Nikhil and Darius got the chance to see their solution come to life as they spent a month implementing their winning solution with the AirQo team.
“This was our first competition where we have supported the winners to help implement their solution, and we are so happy with how successful Darius and Nikhil have been in helping AirQo implement their models, ’’ says Celina Lee, CEO of Zindi. “It means a lot to us to provide real-world work experience for our winning Zindians, like Nikhil and Darius.”
“The most fascinating thing about the AirQo challenge was the ability to implement our winning solution while working with a team of experts at AirQo to predict air pollution, and directly make an impact in people’s lives,’’ says Darius, of his once-in-a-lifetime opportunity with Airqo. “One key achievement that I am most proud of is that now I am more confident in managing a machine learning project from start to end,” he added.
Real project experience
Paul Green, Project Manager at AirQo, noted that although data science is a growing field, it is not widely offered at local universities and many individuals need to teach themselves through such competitions.
“This is great, but it is often hard to get experience on a real project and to see the model through to completion in the real world. I hope we have been able to help Nikhil and Darius through that process, and provided valuable experience that will help them in their careers; not just as competition winners but also with real world work experience.”
Until recently there has been a lack of data on air quality across sub-Saharan Africa. The ability to accurately predict air quality over short time periods using AirQo low-cost network of sensors will empower everyone from governments to families to make informed decisions to protect health and guide people’s actions.
“Darius and Nikhil’s air quality forecast solution will help AirQo identify serious spikes where pollution levels are higher than expected, investigate the cause and try to minimise it in future,” notes Green.
Professor Francis Pope, DAQ EA Project Lead, noted that their aim was to not only increase the number of people who have the skills to look at these challenges, but also to improve our understanding of where air pollution hotspots are. “The competition was a great success: now we have a better prediction of where hotspots of air quality problems are in Kampala, Uganda, and we hope we can extend it across Africa in future.”
This article was written by Maclina Burungi and first published on the AirQo blog.