Transcript: Project Beacon - Jackson Wilkinson
Hi, welcome to another episode of Django Chat, a podcast on the Django web framework. I'm
Will Vincent, joined by Carlton Gibson. Hi, Carlton.
Hello, Will.
And this week, we have Jackson Wilkinson to join us, who works at Project Beacon, which
does COVID testing here in Massachusetts and has a long history in Django. Hi, Jackson.
Hey, guys. Good to be with you.
Hello, Jackson. Thank you for coming on.
So I think we first met with Django Boston because Path.ai, which is a pathology Django
company, was hosting the meetups and you were in charge of that space.
So nice to see you virtually.
Good to see you virtually.
Yeah, it's one of those weird situations where we actually met in person first in software
as opposed to, you know, corresponding online first.
And now here we are in COVID times and it's mandatorily virtual.
So, you know, funny how things get mirrored up.
But we live near one another in Boston area, and I saw you at a playground, didn't get a chance to say hi.
And it got me thinking, oh, yeah, what I wonder what he's up to.
And it turns out you're deeply involved with Project Beacon, which we'll talk about, among other things.
But maybe we so maybe what's Project Beacon, right?
That's the most timely thing.
And then I'd love to get your backstory of getting into Django.
Sure.
Yeah. So Project Beacon, I've been with since late June of 2020. And it's a really interesting
organization. It was started in the spring, funded partially by GV, formerly known as Google Ventures,
and F Prime, which is sort of a similar venture arm for Fidelity, as well as the Broad Institute,
which is a major genetics and genomics lab affiliated with Harvard and MIT. The three
of them got together in the spring and said, we need to increase the capacity and availability
of testing in Massachusetts here in our home neck of the woods. And they started Project Beacon,
which is not technically a nonprofit. We didn't file the paperwork, but we're chartered to not
make a profit and basically do everything we can to drive down the cost of testing in this area.
And so, you know, we basically at this point do a handful of different things. But early on,
my mandate coming in as as chief technology officer was to figure out how to duct tape
together a bunch of vendors in the right way to start covid testing programs. And we figured out
that that was going to be a really bad experience for everybody. And we needed to become a software
company, which is not what we expected to do. And interesting. Yeah, I thought that was, you know,
because they, you know, you have a software background. I thought that was it was it was
really, you know, the thought was there's enough software out there that we should be able to sort
of piece it all together. And we need, you know, one or two or three software people to make that
happen. And it was just not going to be able to do it. And think again about the thesis of driving
down the cost of testing. We don't own the lab. So there's a lab price that we kind of can't
control. There are all the vendors that are associated with swabs and tubes and all the
other things that go into it. And we can't control those either. But there is the human cost of
testing and the efficiency of having somebody who's there and staffing a site and puts a swab
in your nose and collects it and brings it to the lab and so on. And we thought that technology
could drive down the amount of time that it took people to test you, and that could drive down the
cost. And so we now operate six large, high-capacity regional testing sites here in Massachusetts,
each of which do, you know, thousands of tests a day for free, paid for by the state.
But in the background, we are driving the cost down by making it the type of experience where you get from check-in to check-out in three to five minutes.
You pre-register online. Nobody needs to do data entry and make errors at the site.
You show up with your QR code. They scan the QR code. You get a swab in your nose.
You head off and you get results in less than 24 hours.
All, you know, in an automated fashion. So there are sort of no data error issues along the way. And as a result, the cost of testing in Massachusetts, the cost of an individual test has gone from over $100 to, you know, generally speaking, you can pay $50 or $60 now, but we charge $26.50 in some of our programs for a PCR test with 24-hour results.
and that drives everybody else to lower their prices as well and uh we're now uh you know
pretty much the state with the most robust testing uh infrastructure in the u.s and we've been doing
testing elsewhere and and we also do testing uh in other ways like uh antigen testing and
this cool technology called pooled testing where yeah my school is starting that this week or next
week exactly i'm pretty sure your your school system uses project beacon for uh oh yeah the
testing site is is right there so i'm in brookline which is just outside of boston so they've been
using it i've gone to the with my family the revere um state testing the framingham one um
we probably don't need to get get into what a disaster testing has been but it's been
everywhere i've seen in you know a solution it's been project beacon can i can i drag you back up
a sec jackson what's the pool testing like what is that that sounds really cool cool testing it's it
it is really cool. It's, uh, so let's assume that a, that a COVID test costs $25, uh, you know,
at its cheapest for one person. Uh, but what if we could put 10 people's swabs into the same tube?
If that tube comes back negative, then, you know, all 10 of those people are negative. And suddenly
rather than costing $25 a person, it costs $2 and 50 cents a person. So we do, we do this for
schools. You don't want to do it in a place where there's lots of COVID, but if you have, if you
have high confidence that very few people in an organization like an employer or a school or a
college uh have are likely to have covid then it's a great way to do sort of weekly surveillance
testing and uh so we do that in massachusetts for hundreds of schools now and it's uh three dollars
per as low as three dollars per person per test that's amazing that's amazing i mean you've talked
a lot about the price but the thing that keeps coming into my mind as you're talking about
bringing down the price is that of course you're increasing the capacity to do the testing as well
And that strikes me as, you know, money, money's kind of fungible, right? It can, they can come up with that. But the ability to get the entire population through a testing program in reasonable time, that's, that's something that we're still coming up with.
And it's an interesting world that we're living in, you know, in these months, since, you know, we have something like 20, 25, 30% of the population here in Massachusetts has gotten a vaccine.
And so, as we look toward May and June, there's a situation where the vast majority of people, adults perhaps, have gotten a vaccine. Kids won't have been vaccinated at that point. But if we have robust infrastructure at schools to be able to do pool testing easily, inexpensively, and that exists at wealthy schools, less wealthy schools, etc., then we have a really good system in place where we can monitor and make sure that new variants aren't creeping up.
and uh it's it's it's a pretty robust look at um at the state of of health here in massachusetts so
yeah we're pretty psyched to have been a part of it yeah and the point you just made as well
about it being available in in all schools not just like the wealthy ones that can afford the
testing like that's important that's right so carlton my schools are my kids go to the public
schools here and next week they're opening it up to grades um two through two and up um so that's
partly why they're doing or that is why they're doing the pool testing to coincide with okay with
that opening up for uh full-time they've had a hybrid i mean every school is doing a little
differently but my my schools have had a hybrid um the last few months yeah same i'm the next town
out in in newton and we're uh i think april 5th it goes back full-time so you're looking forward
to that yeah i see the little shuffle in the chair that you're like you can't wait for that
so carlton i mean not to make it all about covid we want to talk about the tech but like
you're in spain spain was we were hit hard much more hard at the beginning heavy and then we had
a massive lockdown because it was it you know um the initial phase it was italy and then spain was
the next worst affected country and so they locked down very hard over the before the summer last
this time last year and that worked really well until the summer and then of course they opened
up over the summer and all the tourists came from all over europe and there was it was a particular
genetic variant of COVID or of coronavirus, which is detectable all over Europe because,
and it came from Spain, it came from the Costas. It's like, you know, people bought it and they
mixed and took it back home. And so, but they've managed to keep this, the vaccination situation
in Europe's not great. They've been a bit slow off the mark. I think the UK and the US and Israel
have got off much quicker and the EU's been slow off the mark. But I think that's part of it is
about um you know the industrial pipeline ramping up to produce the vaccines in the quantities for
you know the population of the world so i you know we just said a year ago that we where we are i
don't think you would have complained complained i just think you know people in europe are feeling
oh why have they why have they got it and we haven't i think that's the thing but we've had
the schools open the whole time that's the the main thing so my my um we got four three of them
are at school one's off because there was a case in her class so she's off now until you know and
until after easter because um if a case comes into an hour they shut down that whole classroom
sorry they shut down that classroom um but the schools have been open and so each of my children
have had a week off or two weeks off celeste the the one who's off now she's on a third time of
being off but generally speaking they've been there and that's been a you know an absolute
blessing because i don't think i could have taken another six months of them not at school that was
just you know 2020 was pretty fierce so anyway that was all cool that was a cool project beacon
This is Django chat.
Does it use Django?
We do, we do.
We are a Python shop for sure.
And so, you know, I think when we started this up
and decided that we needed to build our own software platform,
it was, we knew we would need to do custom stuff.
This was not, you know,
you couldn't grab off the shelf libraries for COVID testing.
You know, that's not the type of thing that exists.
So we needed a lot of flexibility,
but at the same time we needed to work quickly.
The difference between shipping something two weeks earlier might have been thousands of people's tests, right?
Like we're sort of racing against the clock in a public health situation.
So in July of 2020, we started kicking this off.
And as Will knows, I've spent a lot of time with Django over the course of my career.
I'm very familiar with it.
And in talking to, you know, obviously didn't have time to hire FTEs, full-time folks, and I'm not really sure who would want a full-time role knowing that as soon as COVID is done, who the heck knows what's going on with your job.
You know, I was willing to do that, but I don't know what everybody is.
But, you know, so I started talking to an agency that I'm friendly with because I'm an alum.
It was one of my first jobs after school.
I worked at this agency called Vigit, which is great.
They're historically like a Rails and React shop, but they've done some stuff with Django.
And obviously, you know, those two frameworks are, you know, cousins of each other in a lot of ways.
And so I had a quick conversation with them, said, hey, do you guys think you'd be able to support me?
And this might be a four-week project.
This might be a four-month project.
Who the heck knows?
And we got to work really quickly.
And, you know, by the end of August, we were getting ourselves into gear and had some early launches.
And that test site in Revere, Massachusetts that Will mentioned earlier was our first one to open up in terms of high-capacity test sites.
And we did that in October.
And pretty quickly, we were doing, you know, the plurality of public testing in Massachusetts before the end of the year.
So it was, uh, you know, a pretty rapid pace and, and Django was,
it was the key part of it, right? Like it was, it, we had,
we had to build quickly and we needed to, to,
to do it in a way that was, uh,
where new members of the team could pick it up, uh,
easily and, and get a handle on what we were doing. And, and, and so it's react on the front
end, uh, a graph QL interface to, to Django on the backend. Uh, and, um, and then we, we hosted
on aptable, which is a, uh, a hosting provider that's built on AWS. It's kind of like Heroku
for healthcare, I guess is like one of the easiest ways to describe it. So sort of a, a get and
docker-based uh deploy process okay uh that again i've been using for years on on most projects and
um and so i'm really familiar with it so it's it's really like what's the hammer in my tool belt and
can i force other people to use those those same hammers along the way yeah but i i mean you know
to have a reliable go-to tool chain where you're like you know it's this and when someone can't
you talked about getting a new person on to use the same tools when if you can onboard them say
look it's just vanilla this it's vanilla that it's you know if you need to learn this library
it's just read the read me and that you know it's it's not complicated it's not you know i haven't
done anything homebrew right right right and you know we've gotten to the point where it's a it's
a pretty hefty and complicated product at this point but you know early on we we wanted to you
know just get things moving and and be as flexible and and and fast as as we could possibly be and
you know there there are things that are not always perfect about django but i think it's a
really good swiss army knife for a lot of these types of things and and can go a long way in a
hurry and you talked about um people entering their own data and getting their qi coded so
that's all like a probably a mobile friendly website would would react yeah it's um so it's
a responsive uh web app that uh so we we don't have a native app in any app stores or anything
like that that was sort of a constant question that we got early on but um i'm a pretty big
believer in in the open web and you know we this way we can serve anybody who has a modern browser
at their disposal uh whether it's probably work on a smart tv exactly yeah and um we also were
translated into seven languages so uh that's you know something that you don't often have to do for
you know an early stage uh product here but um we needed to manage that kind of stuff as well so
uh it's uh accessibility is pretty pretty huge to us uh we want to make sure screen readers have
you know a as close to a first class experience as they possibly can so all those types of things
you have to worry about early on in public health situations yeah because you've got the i can't
remember the the code the 503 code is it where it's kind of for any any sort of government thing
it's legally mandated that it's that that's that's true in this case this isn't technically
like a government um website so we don't have those specific types of restrictions but um
we would hold ourselves accountable to to those types of bars anyway it's just you know we want
equity to be a key component of what we offer here so fantastic you're both married to doctors
i think that's part of what informs knowledge on these things i'm really curious to ask so we've
had a number of consultants on, um, from your side, what was the, you know, so you already knew
Vigette, um, but what is that process like? So presumably you said, okay, here's what round one
looks like in terms of functionality. Um, can you talk about on your side of how you interact with
them? Um, and then it grows from there. Like you said, you know, we need login, like, like why
GraphQL, for example, like how are those decisions made? I guess, is it all just you saying, hey,
I want this or how much back and forth is there? There's a lot of back and forth. So I would say
it's probably a little bit different from the normal client vendor relationship in that every
day, you know, at 1030 a.m., we have our stand up and I'm in it. I lead it. And, you know, so so
unlike probably most client relationships in this context, it's very, very hands on. They're
full teammates of mine. And we interact that way. But when it comes to decisions around
architecture and tech, I think in the early days, it was often me making a proposal and seeing how
that stuck with the team. But GraphQL, for instance, I'm probably more of a GraphQL skeptic
than i am like a graphql fanboy uh and you know i think that there are uh really nice things about
it it's really flexible uh but at the same time if if i'm this is how old i am if i'm looking at
like web logs to get a sense of like how certain features are used it's just a bunch of post
requests to graphql like i can't see anything right like it and and so you know the the ability
to to track and measure certain things carlton feels that you're the youngest one on this call
by the way but you know i'm like looking at nginx logs and i'm like wtf what the heck is all this
stuff and you know where where restful stuff you're like oh these are the features that are
most used right i guess as a product person and as a as a sort of exact um you know graph ql doesn't
doesn't make my life easy but uh on the other hand you know it really came down to speed and
being able to run it's faster right right and and since speed was was going to win here it was
hey we we can really iterate on this and and expand our our scope in this api a lot faster
with graphql than we'd be able to do with a restful framework and i was like all right that's
a convincing argument let's do it um and i still complain about the the measurement and you know
logging uh abilities there but but on the whole it it has been uh i think the right decision from
from sort of a speed and uh and effectiveness front there so a lot of that was give and take
um, you know, there, there were, we sort of had, again, I'm, I'm an old man. And so, you know,
single page, uh, front end frameworks are occasionally tough for me to swallow when it
comes to, you know, the challenges of how those get debugged and, you know, not necessarily having
a full request associated with each thing that you want to do and so on. There's a double-edged
sword to all those things. And so, you know, we have those conversations and ultimately I want
the team to be able to do something that they're comfortable and confident in. And sometimes that
means I'm going to bend, but other times I need to push the team to take on some things that
might be a little uncomfortable for them because either, you know, we need good audit logs of
what's going on because this is healthcare and, you know, those types of concerns might ultimately
just need to, to win out. And so it's, it's been a really great collaborative, um, relationship
that, that we've had over these past, you know, eight months or so that we've been working together
and, uh, but it, but it's different from, from your normal, uh, client vendor relationship.
And I think at least from my experience, so it's, uh, it's been really great.
How do you test that? I mean, not just the software testing, but, you know,
because it's healthcare, like, you know, when I'm in line in Revere, you want it to work.
is there any way to to test to do a dry run like with the physical the online and the offline
how does that work i mean certainly early on we were doing lots of dry runs we we have a great
relationship with our lab which is which is the broad institute uh and so you know we have a we
have a staging environment with them as well so we we can you know end to end uh throw stuff through
this but you know i'm folks are presumably not seeing video most of this but i have like barcode
scanners sitting on my desk and, you know, label printers for all sorts of stuff. So a lot of the
folks on our team similarly are sort of equipped with everything that you're going to have at a
testing site, which, you know, iPads and MacBook Airs are sort of the hardware that we use there.
We have these consumer grade Dymo label printers. We have these, you know, Amazon's
bought $60 barcode scanners, nothing is like special or, you know, super hard to come by.
And so we can, we can put them in the hands of, even in this sort of COVID everybody is remote
world, we can, you know, have folks on our team who have access to the same type of equipment that
we would use in Revere, the testing site. But then, you know, when a testing site is going up,
then we spend, you know, a day in advance, running some folks through it and making sure that
everything's working. But in the end, it's just like, keep it as simple as possible so that we
have as few things that can go wrong as possible in the end. And most healthcare systems are
overcomplicated. And in our case, we only need to do COVID testing. We don't need to do
other genetic testing. We don't need to do flu vaccines or anything like that. If you show up,
we know what you're getting. You just get into one line and we need to do exactly one thing right. So
that gives us a lot of flexibility and a lot of freedom and and helps us do a good job of it
can i just follow up there must though have been in production errors so how are you tracking those
and sure so we use we use sentry to log any exceptions um you know familiar tools right
you know it's the best thing out there totally totally i think uh it's it's the clear it's it's
clear go-to in any of these events. I think in our case, we don't have things like Google Analytics
or anything of that nature on there. We only collect the information that you have shared
to register with us. And the only information that you share to register with us is what we
are required to report for, you know, like state public health reporting. So we often don't have
a lot of information uh about you know the users who are going through our product and that
sometimes makes it a little more tricky than it might be if we had a you know if we were a more
typical consumer platform that might have a robust set of javascript based uh you know tracking tools
yeah and we don't have those but uh we can we can generally do what we need to do
there are some challenges one thing uh speaking thinking about europe uh carlton uh we obviously
focus on massachusetts and and more generally new england the northeast u.s and um we do some
testing for businesses uh that want to test their employees and one of those uh employers
has is a multinational uh corporation that has employees in europe who occasionally travel to
the u.s and they need to schedule testing and we found out that if you are on the east side of uh
of the utc line uh we have a front-end bug that just like totally borks the process for getting
an appointment like good and we had no idea that this was the case because none of us were
east of the utc yeah you built this in boston right totally and so we we started getting these
complaints like do we have restrictions that mean people from the netherlands are not allowed to book
appointments i'm like no we definitely don't have those restrictions and we're like it all seems to
work for us and then i changed my time zone and suddenly you know it all blows up so uh the time
zones are always the thing that bites you right yeah yeah like i've got the do you remember
dastardly and mutley i've got a picture of mutley in my mind now for russian russian russian russian
time zones totally um i want to ask about path ai do you have more questions carl no i mean that's
just amazing i think we should swing all the way back to the beginning though and do it you know
and ask you know okay so that's what you're doing right now but how did you get going jackson like
what was your how did you get into programming and django and all those things that's your origin
story yeah and then we can work all the way up to path ai uh sure so i was a a nerdy kid in a lot
of ways and i think there are pictures of me programming in ti basic when i was like i don't
know three uh like a sort of ridiculously young age i think part of me learning to read was
reading these texas instruments ti 99 4a ti basic books and like copying them in onto the computer
and trying to see if they would run and then making changes to them to and you're like daddy
what does syntax error mean and my parents had no idea they had had
no clue. My parents are totally non-technical for at least most of my life. And, you know,
then eventually I had like a couple of neighbors who were mentors who would sort of show me the
ropes. And eventually I had, you know, a four color CGA display where I could, you know,
do more interesting things. And it was pretty amazing and intense. And so, yeah, I was sort
of like my life was baseball and computers for you know most of my childhood um and then i kind
of put it away in college because it was sort of the dot-com bust uh immediately prior to me going
to college and the the impression of people who worked in technology at that point was sort of the
guy who lives in his parents basement and spends his days and nights in his underwear and and and
that kind of stuff and and i didn't see myself in in that role at the time uh this is pre pre-covid
this is this is pre-covid yeah what goes around comes around right now we're all that yeah um
yeah exactly exactly but uh so in college i was uh i was a music and philosophy double major so
entirely entirely non-technical uh carlton has a phd in philosophy so that's why he's nodding
Nice. Nice. I was a fan of, of, uh, David Lewis's modal realism at the time and all sorts of, uh, you know, harebrained ideas. Uh, but you know, even, even through college, I, I, I worked in the network operation center and was, you know, building websites for faculty and that kind of thing.
so I sort of stayed technical but I never took a CS class um and then uh and then after school I
I took a job at a at a public affairs firm as like an entry-level web dev they're like you have no
credentials do you know what you're doing I was like I think I know what I'm doing we'll find out
and and yeah I was I was in DC for that uh Washington DC and um ended up working on some
some pretty great sort of public affairs, uh, campaigns. After a couple of years, I took a job
at Vigit, um, and started their user experience program. So I kind of got a little bit more into
design at that point. Um, and I think I might've just gotten a ding. Let me put myself back on
do not disturb. So, uh, yeah, I took a, I, after that I took a job at Vigit and, uh, started their
user experience team. And that sort of took my life a little closer to design. There are a couple
of funny stories about how I sort of got into that relating to, you know, at one point at this
public affairs firm, our designer had quit and we got a request to do a site for, you know, I think
uh hillary clinton's uh senate campaign uh way back in the day and we had to come up with some
mock-ups and there was no designer around and so we were like what are we going to do these are due
on monday it's friday or something like that crack the knuckles fire up photoshop i'm off
exactly i'm like hey guys you guys keep looking for a contractor i'm going to see what i can come
up with if we like my stuff then great and uh we ended up using it for the pitch and and i think we
we won the the contract and and so i sort of became like the de facto star was born
right right exactly in the office he's like who knew who knew and i mean looking back at that
stuff i would totally cringe but you know it was it was it was nice to be able to to sort of start
to be a generalist at that point and um so yeah at bigot i started the ux team and maybe wrote a
total of seven lines of code during the entire time i was there i was definitely a a designer
And then I went out west to say I want to be at a startup and went to LinkedIn.
And by that time, I was maybe the 250th employee.
So it was not a startup anymore.
And I didn't know the difference at the time.
And it was definitely in growth mode.
I worked on a bunch of features that still live to this day.
I was, uh, you know, partially responsible for profiles on LinkedIn homepage and a lot
of, uh, you know, people you may know and, and, and things like that, that, uh, still
continue to live.
So that was a, that was a cool experience, but I really wanted to be at a smaller, uh,
company.
So the, the CliffsNotes, or I went to Posterous, which for those of you, uh, who've been around
a while was this rad platform that competed with Tumblr in the, in the early days, but
you could, you know, post your, post your blog by email or via the web. And, uh, I remember that.
Yeah, it was, it was a pretty popular, uh, platform for a while. And, uh, we had a great team
in the mission in San Francisco. Uh, so that was a lot of fun. And then I sort of married
into healthcare and started a, a digital health, uh, product called Kinsights, which is a community
for parents of kids with rare and complex diseases that I was the CEO.
And then I demoted myself to CTO and worked on that for four years.
And we sold that to Care.com, which brought me to Boston.
Care.com is a service best known as a place that you find nannies for your kids.
But you can also find housekeepers and senior care folks.
And so I ran product and user experience there.
And that was, it was a publicly traded company.
So a lot of different sorts of skills being on the, on the management team at a company
like that, then at a, at a small, you know, digital health startup.
And then that led into Path AI, where I spent three years helping build the product design
and engineering teams as we tried to implement machine learning and computer vision to help
detect diseases like cancer and, and other sorts of things like that in pathology imagery.
so that was a super rewarding time and i know that there have been folks on on the podcast who uh
who've worked at path ai as well so uh probably heard a little bit about that in the past so but
that was a great experience and and i took that right up to covet times wow so you've been literally
i mean like you've been over the whole country like you've been west you've been east you've
you know yeah i i mean haven't spent a lot of time in the south uh at this point it's the coast
carlton's just the coast that is the whole country no there's what's in the middle boulder boulder
can i ask though i want to ask about um so at django boston you gave a talk on a history of
django um which i know you said you threw together but i found really interesting going through each
of the major versions and the features when did you first see django because you were like really
early on the django train as i recall yeah i think my first django project uh was during that that
very first job of mine, which would have been in 2005, 2006 sort of timeframe. So it was pretty
early on. It was pre 1.0. The magic removal branch was the big news in Django land. And
I had a little bit of Python under my belt and it was spending most of my sort of professional day
in PHP and, you know, sort of wanted something that was a little more MVC-ish and was looking
around for what to do. And it was like, all right, well, Rails is pretty popular, but, you know,
I'd have to sort of learning curve up on Ruby, but there's this Django thing. Let me see. Let
me play with that a little bit. And it, you know, it did what I needed to do for those early
projects and um it just started becoming uh becoming that hammer in my belt and fortunately
it i remember having discussions at the time about this you know fly by night nobody's ever
heard of it uh framework why would we ever use this on on client projects uh and you know as
time went on uh that that decision sounded a little less crazy over time so it was uh it ended
up serving me well so i'm pretty fortunate in that regard that's lovely because like 15 15 years
later like django's sort of everyone's like it's too old now we couldn't possibly use it now it's
too old it's boring it's like you know it's i know it's exactly why you have to use it but
it's a different topic exactly exactly uh so i want to i want to ask about path ai a little bit
because we did have we had uh robbie groden on but i mean that was at a i'm trying to remember
did you you switch them over to Django right they were on something else when you joined
because you joined well you were doing wearing a lot of hats right you were product and head
of engineering yeah something like that product design and engineering for for most of my time
there as as we grew that sort of shifted uh but yeah there were a lot of hats that that were worn
and early on when it wasn't it wasn't Django yeah right that's right it was yeah it was it was Java
Um, and we had, we had a couple developers who were focused on the web.
Most of the developers on the team, most of the engineers on the team were focused on
machine learning and they were all, you know, very heavily Python based.
And, you know, we, we had the, the folks who were working on the, on the website of the
platform, uh, we're not, they were experienced engineers, but they weren't super
experienced web developers. And so they weren't super tied into the Java frameworks that they
were using and weren't super familiar with it. And there were a few sort of infrastructure choices
that made it, that made most of our use cases a little bit more edge case. And so the documentation
was a little more challenging and, you know, whether we were, because we were in Java mode
rather than Scala mode, and so most of the documentation sort of assumed that you were
operating in Scala, and we weren't, so the documentation was a little sparser, and we
would run into these edge and corner cases seemingly all the time.
So as the team grew and as we were making decisions about where we were going to be
going and what we wanted the future to look like, it felt attractive to me to be able
to have all the engineers at the company have a lingua franca and be able to use Python more
broadly and be able to sort of borrow from the work that everybody was doing. And that meant
that Django became a potential option there. So we had a lot of conversations with the engineering
team about directions that we could go. I piloted out Django on a couple of small micro projects to
to help folks get accustomed to what that would look like and what that infrastructure felt like
without having to sort of overhaul the system. But we had, we had a couple of, you know, smaller
services, some, some internal, some external that, that were that I just rolled up my sleeves and
implemented in Django and folks were like, oh, huh, you can do that pretty easily. And that was
pretty fast. And, you know, oh, that seems pretty nice. And after, you know, a few weeks of that,
or perhaps a few months of that, we decided that it was worth it to sort of deprecate
the Java side and start to spin up the Python and Django side.
So a lot of Django REST framework, and in that case, it was more of a RESTful API initially
than GraphQL.
But, yeah, as Robbie talked about, you know, it became a pretty robust and pretty large part of what we did there.
And we definitely did get some mileage out of allowing people to sort of flex between working on the customer-facing product and working on the machine learning platform, which was also in Python.
So it was nice to have that sort of commonality across the org and the ability to share some work and some libraries here and there.
So and I and I recall you were you you all were very current with Django editions because Django versions, because this is we talk about this a lot in the show, you know, in part because you had started early.
But I think also just for you, like, as I recall, you're it's pretty important to you to stay up to date with the latest versions on your projects.
it is i you know i was afraid you were going to ask me about that because i'm worse about it right
now with project beacon than i think i've ever been in my career but you're running off master
we're we're definitely you know at project beacon i think we're um you know maybe like
i think we're 3.0 somewhere i'd have to even look like i'm like that disconnected from it right now
mostly because we you know don't always have the time to do all the hygiene things that we
would totally love to do um sure but i think a a well run and well executed product uh that does
have good hygiene does keep uh up to date on those types of things they're they're the obvious reasons
like security uh patches and bug fixes um but also just that um you know the the features that
come along the way are good and useful and there will be useful improvements and i'd rather have
small projects to get us up to speed and up to date on a regular basis than have the enormous
project when you find out that, you know, suddenly you're two years behind and it's going to take,
you know, three months of engineering time, all hands on deck at high risk to get up to a more
recent version. So, you know, I'm a pretty big believer in that. And we did a pretty good job
with uh at path ai and i think we're we're we're still okay it's only a it's only an eight month
old platform so it can't be that terrible but uh we're we're a little further behind than i would
like i think okay that's good how do you find it from a um from a hiring perspective because i
always think you know if you if you're on the latest version it's much easier sell to a new
recruit and you know we're on you know one point yeah i think that's i think yeah exactly i think
that's part of the pitch. If, if, if we were saying, yeah, that we're in the 1.x series or,
Hey, we're still crapping along on 1.8. That was at one point an LTS version, maybe, um, you know,
that would be, that would be a tough situation. And, um, and I've seen at other companies that
just be a slog and be really challenging. And so, you know, hiring it's, it's an important
component um security and and compliance i think it's an important component uh on one hand you
think of compliance as like pinning a version and having something reliable and few changes but
uh it just exposes you to to every vulnerability that comes across and you know makes it makes it
a much bigger uh decision process so uh but yeah engineers are generally happy when they can go and
hit latest in the docs and you know actually use them so it's it's it's good to be in that position
i find your compliance point is a good one i always like i always struggle with how a company
can justify saying you know we're using a version of software which has been declared end of life
by the people who maintain it like for just from a liability perspective you know if there was some
bug that happened because you hadn't patched a version and then you got sued for that and you
wouldn't have a leg to stand on because there's no reason to be on that software well you can hire
red hat now right okay your python 2 for you done all done uh yeah that um i that i to your point
though on versions i think for people who are interviewing at places that's one of the best
questions to ask no matter what they say just say what version of django are you on if you're hiring
for django that tells you what you need to know they can talk all they want if they're on one
five it tells you it definitely tells you something probably yeah it doesn't tell you
everything but it's a it's a pretty good tell i mean and i again i be partly because you were
hosting at path i was hosting the django boston i was surprised that you the company was so up
to date because it is unusual because it is hard and it requires a deep commitment at the top to
to stay on it right it's a better experience to as an engineer it's hard to justify if you're
sort of breaking it out like as an exec saying like yeah we spend yeah i don't know whatever
five percent of our you know time on you know upgrading libraries or something like that that
doesn't sound really great but if you make it a seamless process and you do it so that there are
very you know few changes that need to happen along the way because you're constantly up to
date then nobody notices and you get all those those positive advantages as well so um you know
it's definitely, I think that you either need to stay very up to date or you will inevitably fall
very, very far behind because those, that middle of the road, Hey, this is going to be a midsize
project for us to, to update is, is probably the toughest one to justify. So, um, so yeah,
I tend to, to err on the former side. So I like that. So what, what haven't we covered? I mean,
we've covered a lot of things, I guess, you know, you don't know, but at some point testing will be,
So Project Beacon, so the code, is there a potential for the code or the company to, you know, who knows what will happen with COVID, but variants may be around for a long time. So presumably the code base is useful in some way.
Yeah, hopefully. We certainly, it hasn't been part of our objective to build this for the long haul. Though insofar as we do think it's important to build it to be robust and scalable. And so a lot of those things end up going hand in hand. So we have a really great test suite, end to end tests that are like Cypress and Selenium, decent, decent unit and integration tests along the way as well.
and a good deployment pipeline and CI pipeline there.
So those types of things we've spent the time to put into place
so that we can be confident that when you show up at Revere
and you've got a QR code to get tested, that it's actually going to work.
The long-term and the legacy of that is something that we sort of continue to talk about.
We sort of have our hands full with actual work.
But, you know, every month or two, we kind of take stock of that.
And it'll be interesting to see what the timeline looks like for this.
I think that's partially dependent upon the effectiveness of vaccines against these other variants.
We know that the tests that we run can detect these other variants.
So it's, you know, certainly in the during 2021, it'll continue to be important.
And Project Beacon will continue to, I'm sure, do a lot of testing throughout the rest of the year.
Uh, but what things look like in 2022 is, is tougher to forecast and, uh, you know,
we'll, we'll have to see what, what the world looks like as we get closer to it.
It would be, we, we certainly all want this to have as lasting an impact as it possibly
can.
And, um, you know, we're, we're now doing a lot more testing outside of Massachusetts,
uh, than we, than we ever have before.
And so, yeah, there's a, there's a world in which this can, you know, be a component in,
uh whether it's regionally or or nationally or or beyond that in the in the public health
infrastructure and uh you know hopefully we can make a decent contribution to to that ecosystem
carlton anything else you wanna no no i just i think that's stonking i love it when we have
you know i love all i guess but i love it when we have a like in here's how it goes hits the
real world that's you know yeah and maybe it's you know jango jango seems somewhat common in
in healthcare, I don't know if that's a fair to say Jackson, but you know, maybe partly because
the maturity of the platform and the needs of health tech are different than, you know,
a social media startup. I think you do find it in healthcare. Um, not uncommonly, I would say,
you know, common is, is, is a tall bar there. Um, and it would be, it would actually be like
a really disparaging comment to say that Django is popular in healthcare because
bad software is really popular in healthcare too. So, uh, you know, I think not uncommon is
the way that I would put it. And I think part of that is, uh, because of all the, all the great
things that we've talked about, part of it is also just because of the, the Python ecosystem,
right? Like, um, the, the Ruby and, uh, and, and node, uh, communities have tons of libraries for
web development and, you know, sort of the, the typical tools associated for that. But, um,
I just think back to when I was working on Kinsights, that community platform for parents
of kids with rare and complex diseases, we needed to render out growth charts to help
parents just chart their kids' growth.
And what do you know?
There's a Python library for that, right?
Like in no other ecosystem do you find that kind of stuff.
But because of the scientific and medical research community using Python, you end up
having a lot of these crossovers that make your life a little bit easier a little bit simpler a
little bit more reliable and you don't have to reinvent the wheel all the time in in health care
so i think that's a uh key component there too it's like the the xkcd import anti-gravity
so not far off from that in a lot of cases for sure well thank you so much for taking the time
i know you're very busy these days um so i appreciate coming on to share the story interesting
awesome can i can i make a a little uh public benefit pitch at the end oh yes please yeah do it
so um one thing that i didn't talk about um after i left path ai just i think this week last year i
joined uh the u.s digital response usdr which is a non-profit organization that uh helps staff uh
important technology projects uh especially related to covid and and you know sort of the
2020 world that we're living in. And so I led a team with USDR for, uh, four or five months,
had the opportunity to contribute to some really amazing projects like, um, the, uh, in New York
city that when they had the PPE crisis and they didn't have enough PPE to go around during the
first surge, uh, my team worked, uh, with, uh, eight departments across the city to help them
get a handle on where PPE existed and where they were sourcing it from and help them get a, get a
handle on what that supply and demand looked like. We also worked on hospital capacity planning in
Pennsylvania and outbreaks of COVID at homeless encampments in the Bay Area, in addition to many
other projects that USDR worked on that I didn't have a chance to contribute to. So I think that
for folks who are learning Django and learning Python and wanting to get their feet wet,
or folks who have perhaps a gap in their career
or looking for something to do in off hours
to extend their skills,
USDR is a great place to spend that time
and make a contribution to efforts across the country
that can help with all sorts of culturally,
socially, and politically important issues
and a time well spent for anybody
who is looking to develop those sorts of things.
So strongly encouraged that US Digital Response,
check it out.
Yeah, we'll put a link to that in the show notes. But thank you for that. I should ask you maybe offline. We could probably get someone from there to to come on the show and talk about that. But we can we can do that separately.
Yeah, totally. Now I'm sure they would. I'm sure they'd love to.
So if anyone wants to help Project Beacon, is there any immediate way to help Project Beacon? Or what can someone do?
No. So Project Beacon is interesting because we're not actually a nonprofit. We just have
charged ourselves to not make a profit, but nobody has put in the paperwork or paid the
lawyers to make us a nonprofit. So you technically can't volunteer to, uh, to work with Project
Beacon, uh, for better or for worse. Um, but certainly we, we are hiring for, uh, a lot of
roles, uh, technical and non-technical. So if you're in the Massachusetts area, you know, we
tend to, or in, or in the new England area, there are definitely roles that, uh, that we could use
help on and and we we pay actual uh cash money for that rather than just taking your time so
uh so if you're interested definitely happy to talk to to good motivated folks and uh and our
our partners at bigot are also uh frequently hiring for more technical roles there and that
applies to anybody uh at least across the country uh here in the united states so they're they're a
great shop and do some great work and um you know would be would be great to work with with folks
who uh who join their team as well well thank you so much we are at um djangochat.com chat django on
twitter um thank you again we'll see everyone next time join us next time i could talk to you bye