When Tristan Ford was a graduate student at Johns Hopkins University, the 2015-2016 Zika epidemic loomed large in global public health. With a background in biomedical engineering, he applied Artificial Intelligence (AI) to solve for one of the primary challenges in the Zika response: insufficient vector surveillance. We spoke to Tristan about the company he co-founded and leads, Vectech, and how its AI-powered mosquito identification system is now mapping malaria risk in Africa and beyond. 

How Does Vector Surveillance Support Malaria Control, and Why Can it Be Hard to Get Right? 

Malaria control programs have limited resources, which they strategically allocate to save as many lives as possible while reducing the transmission risk for target populations. But to do so effectively, they need to have a clear understanding of which areas have the highest risk — and that’s where vector surveillance data comes in.

The highest risk area will have the highest abundance of primary vector species, or those that can carry and transmit diseases. Some mosquitoes are just nuisance species that cause itchy bites, but others are really good at transmitting diseases like malaria. Among the most dangerous for malaria are Anopheles gambiae s.l., Anopheles funestus s.l. and Anopheles stephensi — an invasive, insecticide-resistant and container-breeding mosquito now spreading across Africa. Each of these species behave differently, so it’s critical to know which mosquito species are present so interventions can be tailored to more effectively reduce the mosquito population and thereby, lower the malaria risk in that area. 

The problem is that most malaria control programs cover a large area. Given the cost of transportation and limited resources available, programs may only be able to collect mosquito samples from certain areas once or twice a year, meaning they have limited data available to create a risk profile for the region. In contrast, the U.S. for example, might have weekly data points representing individual locations throughout a region in any given year. Another big bottleneck is the limited number of medical entomologists — the experts who run malaria programs and often examine the mosquito samples one-by-one under a microscope, among their many responsibilities — for an entire country. There simply isn’t enough entomological capacity to run effective vector surveillance programs. 

How Does Vectech’s AI-powered Mosquito Identification System Support Surveillance? 

When I was at Hopkins, we realized that a relatively new technology at the time, computer vision, was capable of visually distinguishing the taxonomic characteristics of different mosquito species — and then automating that process. We realized that if we could build that capability into a product for vector control programs, we could expand vector surveillance capacity. And Vectech was born. 

Our first product, IDX, is essentially a smart digital microscope. Specimens still need to be collected conventionally from the field, but rather than contending with the availability of a medical entomologist, the samples can be run through IDX. Images of the specimen are processed through our algorithms, which can identify over 85 species of mosquitoes. For malaria specifically, IDX can identify nine primary vector species, 13 if we include secondary vectors, and 19 species within the Anopheles genus overall. As more and more partners adopt IDX, they contribute data that allows the algorithm to improve and scale to additional species. 

The models we use are built on what’s called convolutional neural networks, which are really good at identifying the unique visual characteristics that differentiate one class of data from another. These kinds of models are used in self-driving cars and in Facebook’s facial recognition algorithms. Instead of defining what characteristics the algorithm needs to evaluate up front, we use a labeled database of thousands of images of mosquito species and quiz these models over and over again until they pick up the patterns that might distinguish one species from another. This might include factors like color, size, shape, and wing patterns, much like an entomologist uses a taxonomic key to identify specimens. 

How Is IDX Being Used for Malaria Today? 

We work with ministries of health and national research institutions in malaria-endemic countries, including the Kenya Medical Research Institute, or KEMRI, the NIAID International Centers of Excellence for Malaria Research (ICEMR) program in Mali, the Papua New Guinea Institute of Medical Research, the Jimma University in Ethiopia, and others. Our partnerships with each of these institutions includes gathering data to train our algorithms and expand the number of mosquito species it’s capable of identifying. We continue to evaluate how IDX can be deployed in each of these countries as well as the broader regions, while also training it on invasive species like Anopheles stephensi, which is present in Kenya. 

The Walter Reed Army Institute of Research in Africa, which is responsible for supporting the U.S. Africa Command, also uses IDX as a secondary reference tool and in support of disease surveillance for regional military exercises such as 2025 African Lion in Morrocco. When they collect specimens from the field, they use IDX to double check and verify a species identification. They are also contributing data to improve the algorithm’s performance. 

The U.S. military has a long history with vector-borne diseases. Malaria was a huge problem for U.S. troops in World War II, especially in the Pacific Theater. And today, maintaining U.S. military readiness to operate in any part of the world includes managing the risk of vector-borne diseases like malaria. There are very few medical entomologists in the Armed Forces, but they still need good data to make sure U.S. troops are protected. A soldier without any entomological training equipped with IDX can provide that capability. I’m actually in South Korea right now to explore how IDX and our new Scout system can protect U.S. troops along the North Korean border. North Korea is a reservoir for malaria which has led to cases among U.S. troops who serve in the region each year. Essentially, we want IDX to be an AI diagnostic entomologist that can be deployed anywhere in the world. 

What’s Next? How Is IDX Being Scaled?

The next hurdle we’re looking at is, of course, cost. We’re trying to make sure that our products are affordable not only for the U.S. groups that we work with, but the research institutions we work with in Africa, like KEMRI. As we scale, we’re looking to see how we can bring down the cost to increase access at all levels, including at the health district level, for low- and middle-income countries that need access to this technology.

This year we’re launching a new smart mosquito trap product called Scout. By capturing, autonomously imaging, and identifying the mosquitoes in the field, it’s tackling the wider workflow challenge of not having enough people on the ground to go out and collect specimens regularly. The Scout smart trap can collect data remotely, and with greater resolution, to tell us what’s happening minute-by-minute throughout the day with a mosquito population. 

We’re also looking at some interesting research questions on the horizon, beyond species identification — like age grading mosquitoes to know how old they are and sibling species identification, which is really challenging for species within malaria vector complexes. If we can answer those questions, then we can move IDX beyond the capabilities of conventional identification by an entomologist — transforming surveillance tasks that usually require a wet lab or DNA barcoding into something that can be done anywhere, by anyone in a matter of seconds. 

This conversation has been edited for clarity and length.