Good morning,

We’ve hit our stride this October. As well as completing our £100K SEIS raise, we’re getting our very first customers - and making the technology decisions that will make Sundog delightful to use for them, and sustainable for us.

Alex Mitchell | CEO update

I’m currently deep in sales conversations, making sure we’ve got a strong group of alpha customers at launch. Every conversation I have affirms that we are building a product that matters.

Our first customer leads a cancer research group, looking at how cells are influenced to replicate - and how we might stop cells from proliferating out of control. It’s easy to state this as a goal, but a cell in a living system has a huge range of influences. The cells around it, the form of the tissue, and the environment around that tissue can all affect the behaviour of a cell in a biological system.

That’s where the collaborations Sundog enables are so powerful. A cell biologist can identify the interactions of cells with other cells. Pathologists have expertise in the level above that; both the features (like blood vessels) of a tissue they’re looking at, and what healthy and diseased tissues look like. By collaborating with pathologists, who mark up important features and characteristics of the tissue, the biologists who use Sundog will better understand the context in which cells proliferate (or don’t), build a more complete picture of what triggers out-of-control replication - and how we might prevent it.

Collaborations across disciplines are incredibly powerful. So are collaborations across borders. I’ve been speaking to a group leader in Malaysia whose teams are often seconded internationally - to their university’s main campus in Australia - for three months at a time. That’s long enough to learn new imaging techniques, acquire images and do some preliminary work on them. The difficult part is the analysis, which always happens back at their home campus. By continuing the real-time collaboration, even with the distance, the scientists in Australia and Malaysia can keep working together closely - and everyone can learn more, together.

The scientists to whom I speak are hungry for a better way to work together. They know that the collaborations that advance scientific knowledge for everyone will happen across specialisms, disciplines and borders. We’re building the software that will bring everyone together, exactly where they need to be; in the ground truth of the image itself.

Tom Armitage | CTO update

This month we’ve been working on everything to do with uploading files to Sundog: getting files from your computer to ours, resuming failed uploads, working out what’s inside your files, and preparing the various pages of the file for use in Sundog.

We demonstrated something like that in the end-to-end-prototype. But that prototype is made of string and glue: it’s not robust, it doesn’t cope well with surprises, and adding more string and glue is not the way to fix it. This time, we’re focusing on building something that will get us where we need to be.

That means thinking in more detail. Thinking about what happens when things go wrong, how we can reliably communicate between our web application and image processing containers, how we can communicate state to the users, and how we can make this as performant as possible.

Speaking as Sundog’s CTO (rather than as my other roles of engineer or designer): there’s high value to us in doing this right. The details of engineering problems directly connect to the user’s experience of the product, and even to its business model. The improvements we’ve made to the upload systems now should pay off multiply for us down the line.

For every file a user uploads, we generate quite a few more: the SZIs (image tiles) we make of the layers you want to work on, thumbnail jpgs, metadata JSON, and so on. To keep things simple, we plan to bill labs just on the volume of files they’ve uploaded. If you upload a 1.2gb file, that counts towards 1.2gb of your quota.

There’s a magic number I think about quite a lot - let’s call it n - and n is the ratio of how much we have to store compared to what you’ve uploaded. For instance, if every 1.2gb upload generates 0.6gb of extra files (SZIs, thumbnails, etc), then n is 1.5 - we store 1.5x what you upload, and we owe AWS for 1.8gb of files stored (as well as all the charges for moving them around).

The smaller we keep n, the better price we can give our customers, and the more wiggle-room around margins we have when a project turns out to be particularly large.

In our summer prototype, n was very high. Each uploaded tiff file was split into one tiff per layer, and then into tiles (made of individual tile files, in the pre-SZI days), and that’s before you think about figure outputs or metadata.

Our current codebase works hard to keep n low. We’re now storing tiles as SZI files, which helps keeps transaction costs down. We’ve found ways to avoid having to ever break larger tiffs into small ones; instead, we pull out the data from the original file on demand, which noticeably reduces storage costs (a saving we can pass on to users).

And this approach to file processing directly improves users’ experience. The main thing that made our prototype slow to process files was not the tiff processing itself: it was copying the extra “extracted” tiffs around the storage network. Thanks to the new approach, where we don't ever duplicate tiff data, Sundog now runs a lot faster - in places, it might even not be worth putting the kettle on after an upload. Seriously quick.

That's why the details matter. When I go on a bit about the deep nooks and crannies of the tiff spec, or of our infrastructure for processing them, it’s not just to geek out about technical matters. Those details have a direct, measurable impact on the user experience, on the price we charge our users, and the margin for safety (and profit) we can build into the operating plan. Really, it’s all part of making the best product we can.

Our work this month has shown us the power of spending time in the details. It’s also shown us, particularly vividly, what the product we’re building is going to mean for the people who use it. Thank you for following along with us. If you want to talk with us, reply to this email or get in touch with alexandra@sundog.bio.