The Center for Data Innovation spoke to Eve Tamraz-Najjar, co-founder and chief scientist of SensioAir, a London-based startup that uses home air sensors and an app to deliver personalized predictions for allergy sufferers. Dr. Tamraz-Najjar discussed how machine learning can help people manage their allergies, and the value of mixing data from different sources to deliver specific recommendations.
Nick Wallace: SensioAir puts an air quality sensor in the home and links it to a smartphone app. The utility of a sensor that monitors the air for allergens is pretty obvious, but how does SensioAir personalize that information for individual users?
Eve Tamraz-Najjar: I think SensioAir is a novelty in two ways. First, we have the first sensor that is actually able to identify airborne particles. So far, every air quality monitor that you can find on the market only measures dust. So they can tell you if there are a lot of floating particles or not, but they can’t really tell you what these particles are. If you look at the composition of household dust, you find mainly skin flakes and tissue fibres, but these will not trigger an allergic reaction. What we’ve really worked on is a sensor that is able to identify allergens, like pollen, dust mites, and mold.
Once the user understands what kind of allergens are in the air, he is able to take actions that are more tailored and useful for prevention. The second innovation is that we enable the user to log his symptoms on a mobile app, and we cross-reference the user’s symptoms with whatever’s in the air at the same time. In the long run, we can understand what the user’s triggers are, and give him very specific recommendations when the levels rise. We also take into consideration temperature and humidity levels. Based on all this data we can give very specific recommendations to try to decrease the amount of allergens in the house.
For example, I suffer from allergies, and until now the only thing I could do was take very generic and blind action, like removing the curtains, removing the carpets, not owning any pets, et cetera. But I couldn’t really see the impact of these actions on the air. SensioAir allows the user to follow the efficiency of these preventative steps towards allergy and asthma management.
Wallace: How does SensioAir use machine learning? What factors does it predict?
Tamraz-Najjar: We do two things. We are able to prevent allergen levels based on the time of the year and the environmental conditions outside and in your home. We are also able to predict your symptoms based on your triggers and your sensitivity to specific allergens. The algorithms that we trained are basically using environmental data and symptom logs to link the two and work as a predictive model for both.
The algorithms get refined as the user logs information. The first time you log your symptoms, we might have a higher margin of error than later on. The more information you log, the more the algorithm becomes specific to you. It’s the same for your environment. The more you use SensioAir and let it run in your house, the more we will be able to understand the patterns of pollution peaks or allergen peaks over a week or a month, but also over the year, because these factors are often seasonal.
Wallace: What other data does SensioAir use, besides what is produced by the individual device in the user’s home?
Tamraz-Najjar: We also use outdoor data. For that, we set up our own devices to monitor pollution, pollen, humidity, and temperature, and we also set up deals with governments. For example, Cyrille—my husband and co-founder—and I both have family in Lebanon, and we made a deal with the Lebanese government where they gave us access to their air quality monitoring network, and we can share that on the mobile app. So the information comes either through partnerships or from our own devices.
The outdoor data is available for free on the mobile app, so even if you don’t own an indoor device, you can still check the outdoor air quality, log your symptoms, et cetera. It works quite well for people who are allergic to pollen, because that mostly comes from outdoors. We wanted to give people free access to pollution info as well. For indoor data, we cross-reference indoor and outdoor pollution so we can push specific advice on things like when to open or close the windows.
Wallace: You hold a PhD in neuroscience—can you tell me a little about how you went from that field to creating SensioAir?
Tamraz Najjar: My background is in toxicology, health, and environment, I did my masters in drug design, and then I specialized in neurosciences. I then met my husband at the end of my PhD. We were both very interested in biotechnologies, and we are both heavily allergic, so we decided to find a solution together, and that’s when we started SensioAir. So it was initially a selfish endeavor that turned out pretty well, we realized that a lot of people were interested in this kind of solution, so I decided not to pursue my academic career as a researcher. But I find that having a startup, I get the best of both worlds: I do a lot of science, but I also meet a lot of people and do business. We’re in our third year now.
Wallace: Home AI assistants are becoming quite popular, and there’s a lot of buzz around personalized medicine and the idea of the “quantified self.” Do you see those technologies converging over the next few years?
Tamraz-Najjar: I think that we’ve focused a lot on the quantified self, and then we focused on quantifying the environment. But what really was missing until now was the link between the two. Once we actually implement this in all the different fields, we’re going to have very powerful insights, and that’s what we’re trying to do with SensioAir. We take user-driven health logs and link it to environmental sensors. I’m really eager to see what we can do and how we can push this, and not just in the allergy field. If you take apps like MyFitnessPal, where you have the food log, it would be amazing to link that to the physiological response, and I’m really interested to see how we’re going to be able to merge this environmental and health info. I think this is really where the power lies, and it will allow us to make very informed decisions, and push us towards personalized medicine.