Cleaning our water systems revolutionised public health. The next step is to clean our air, and prevent potentially catastrophic pandemics from engineered pathogens. The question is, what's the best way? Well I tried a bunch of things... One way to clean air is by repeatedly changing it with fresh outside air. But this requires mechanical ventilation systems that are **expensive**. So, we can clean air instead with 222nm light. This is called Far UVC. Evidence so far has shown this to be safe, barring some eye health concerns. Far UVC’s mechanism of peptide bond breakage is already (kind of) known to first order. Whilst far UVC still has unknowns in efficacy, these is progress in meta-studies like Blueprint’s AIR programme. Glycols are in a similar spot too. It's potentially cheaper and easier to use Far UVC to clean air, rather than replacing a entire ventilation system. But the question is, when you get the lamp, where do you put it? How you install Far UVC dictates it effectiveness, and there are a lot of questions. ![[Screenshot 2025-08-04 at 08.32.47.png]] Do you put it in corners? How strong do you set it at? What if there's no electricity in the ceiling? What if there are heaters that makes the air rise? Does it work only if the room is sufficiently ventilated? Right now, researchers model how Far UVC cleans air by simulating droplets exhaled from infected people, and solving the Navier-Stokes equations. The problem is, it's complicated to do this at scale. Building environments vary, CFD assumptions are dubious. Implementation is going to rely on a transparent, open-source and government sanctioned model that UVC installers could can use. **A nice platform is also going to make people more confident in buying it, if they can see an interpretable model in front of them of how theses things clean air.** That's what I'm building! Step one: I tried a bunch of different lidar scanning apps. The best app so far, which I could scan my room with, was MagicPlan: ![[IMG_6535.png]] From the phone view, the app allows installers can also add furniture. Adding furniture is important because we don't want to point our lamp towards people's eyes. From our furniture, we can infer where people will be sitting and looking. This information does feel crucial - for example the top left corner here feels like the best place to put it, if it's corner mounted. You don't want to point it on the same wall as the TV because then it'll point towards people's eyes. ![[Screenshot 2025-08-09 at 08.10.22.png]] Then I used the API to make a 'central modelling' tool for installers in the field - installers can do scans in the field, then send them to a central database. ![[Screenshot 2025-08-08 at 17.18.23.png]] Building file formats (dxf) are kind of hard to use, and we need to make them easier ![[Screenshot 2025-08-08 at 17.44.05.png]] So that we can do lighting simulations like this but in more difficult spaces ![[Screenshot 2025-08-08 at 17.48.05.png]] And then implement simpler, more interpretable models - CFD is complicated - Not scalable for mass usage / not transparent and easy to get distracted with ![[Screenshot 2025-08-08 at 17.39.51.png]]