348: Flop Onto the Bouncy Castle

Transcript for Embedded 348: Flop Onto the Bouncy Castle with Whitney Huang of Zipline.

EW (00:06):

Welcome to Embedded. I am Elecia White, alongside Christopher White. Today, we are going to talk about professional Tacocopters. Well, actually, prescriptions delivered via drones in places where roads aren't viable. Our guest is Whitney Huang from Zipline.

CW (00:26):

Hi Whitney, thanks for joining us.

WH (00:28):

Thank you so much for having me.

EW (00:30):

Could you tell us a bit about yourself?

WH (00:33):

Sure. As Elecia said, my name is Whitney Huang, and I'm an embedded software engineer at Zipline. Zipline is a startup, and we make high speed, fixed-wing drones that deliver essential medical supplies around the world. I joined Zipline a year ago straight out of college. I actually have a background in mechanical and aerospace engineering. I love building robots and I'm really excited to be here today.

EW (01:06):

We are excited to ask about drones and some world travel, but before that, we want to do a lightning round where we ask short questions, we want short answers. And if we are behaving ourselves, we won't ask why. Are you ready?

WH (01:22):

Sounds good.

CW (01:25):

Do you like to complete one project or start a dozen?

WH (01:26):

Well, in practice, I definitely start a dozen, but in theory, I would definitely like to complete one.

EW (01:37):

Do you have a favorite acronym?

WH (01:40):

That's a good question. I don't actually have a favorite acronym, but I have a least favorite acronym, and that's GNU.

CW (01:49):

It's one of those recursive acronyms, it's very annoying.

WH (01:53):

Yup.

CW (01:55):

If you could teach a college course, what would you like to teach?

WH (01:57):

I took this really amazing course when I was in college called Transformations and Art and Engineering, and it combined art and engineering, and we got to do a lot of really cool embedded-y projects and it was really free form and very creative. So I would definitely like to teach a class like that.

EW (02:18):

Drone or quadcopter?

WH (02:22):

I guess quadcopter is a subset of drone, depending on your definition, but-

CW (02:28):

She said fixed-wing, so no longer apart.

WH (02:31):

... Yeah.

CW (02:34):

What is the silliest drone application you've seen?

WH (02:41):

Funny question. I guess one of the summers for my internships, I worked at a startup where we tried to make an indoor drone. It was basically a quadcopter encompassed in a ball and it looked like the death star. That was pretty fun.

EW (03:03):

Cool. Do you have a tip everyone should know?

WH (03:08):

Yeah, don't be afraid to get started. I definitely seen many of my peers get really intimidated by hardware projects coming from, say more pure software background, but don't be afraid to just get moving and make mistakes and learn along the way.

EW (03:29):

Tell me about Zipline. Fixed-wing, launchers, give us the rundown.

WH (03:38):

The technology is pretty cool, but it seems random and strange if you look at it from today's perspective, just watching videos online. How it works is actually we have fixed-wing drones. They look like tiny airplanes. An operator assembles these drones on a launcher, which is basically a high speed catapult. The operators hit a button and these drones catapult into the sky at a hundred kilometers per hour. These drones then move across the sky. And we actually have a service radius of about 80 kilometers, which is about 50 miles. And when it reaches the final delivery site, which is typically a hospital or a health facility in any of the countries that we serve, it will open its payload doors and release a package with a parachute attached to it. That package will glide to the ground and the local doctors who requested the order will go and retrieve the package.

WH (04:46):

Our drones will autonomously fly home, and it will actually do something that we refer to as a recovery, which essentially, it doesn't have landing gear. So what it does is it has a tiny centimeter-long hook at the end of its tail boom. And there is a big truss structure where our original launching location is, so the recovery system is right next to that catapult launcher. And on top of this big truss will be a line, essentially like a fishing line attached across this trust structure. The line will snap up as the drone reports its GPS location at a sub-centimeter level accuracy back to the ground station. And it will engage with that line and catch on. So it's pretty magical-looking.

CW (05:46):

How big are the planes? I have a radio-controlled plane that I have, and it's got a meter wingspan. I think of that as a big radio-controlled plane. This is probably larger than that to carry a payload.

WH (05:59):

Yeah, it is larger than that. It's wingspan is about three meters long-

CW (06:04):

That's pretty big.

WH (06:05):

... and the length is a little bit shorter than that, but it's light enough that I could carry it. I'm a pretty small person, so it was intentionally designed that way to make it easy for operators to assemble them.

EW (06:23):

And the whole hook thing sounds like aircraft carrier, Top Gun-like things. Is that what it looks like, or do I have it wrong in my head?

WH (06:35):

Yeah. No, that's exactly what it looks like and it was very much [crosstalk 00:06:38]

CW (06:38):

But it's catching it mid-air, right? Not-

WH (06:39):

Sorry?

CW (06:39):

It's catching it in midair, it's not trying to land on a runway and slow down and stuff?

WH (06:43):

Yes, that's right.

EW (06:45):

Does it spin around after... It catches the hook and it's still going? How do you make it stop?

WH (06:55):

You could imagine two A-frame structures for the recovery system. And then there's a line in between the two A-frame structures and that line is attached to two poles that swing back and forth. So the drone that comes in-

EW (07:14):

The swing.

WH (07:14):

... it will actually catch onto that line and it'll swing back and forth and that line will pay out and pay in according to the physics of a pendulum to slow down the drone.

CW (07:25):

It's looks like a bat.

EW (07:30):

Just like a bat.

CW (07:31):

Was that one of the... How do I ask this? Was that one of the original ideas for recovery, or what was the process of figuring out how to recover these reliably?

WH (07:45):

There's quite an interesting history to how we came to the solution today. It started off very much like you said, inspired by an aircraft carrier. Since we don't have landing gear, we rely on certain physics and other methods to slow down the plane. And so originally, instead of these tall A-frame structures, we had essentially two fishing poles back to back and with a line across from it, so it was pretty low to the ground. And then after the plane would come in and swoop down low and catch onto that fishing line, there would be a bouncy castle in front of the back to back fishing poles. And it would lightly slow down, decelerate and then flop onto this bouncy castle, big, foam pit, essentially.

EW (08:41):

All right. The bouncy castle. Yes. That's a great landing strategy. Have you ever delivered tacos?

WH (08:55):

We have not delivered tacos. From the inception of the company, they've always really been committed to solving high impact problems. And think it was really a strong, fundamental mission of the company to always have an issue and a problem at hand, and to continuously work to solve those real problems that the early days of our company and our founders were aiming to solve. They did a lot of research traveling around the world to places in central America and Africa. And they would ask people, what are the biggest problems that you face? And medical logistics was actually the thing that came up over and over again. And it seemed like a problem that was quite difficult to solve, but also had such a dramatic impact on society since people were really being affected by just the lack of medical access.

WH (10:02):

And these are people that if they lived in a different part of the world, would definitely survive. We've been really focused on this mission specifically. And there was a while in which our CEO was like, "We would never deliver pizza by plane," but as drone delivery logistics have really become increasingly popular and a lot of companies are racing and scrambling to get this, actually servicing customers, I think we've definitely expanded our scope of what we'd be willing to deliver. But as of now, we are sticking to medical products.

EW (10:45):

And so these are getting the right antibiotic or is this blood or what kind of medical products?

WH (10:56):

When we initially started off in Rwanda, which was our first location that we were stationed in, at the time, their logistics and public healthcare system was really suffering from a lack of blood supply. So we actually started off by delivering almost exclusively blood. But the way our packages are designed, they're a small box, just a simple rectangle, and they can carry things like blood, which are stored in pouches very easily. And so that is what we focused on very beginning, but since we've expanded to different countries and taken on different products, we deliver anything from antivenom to actually vaccines.

WH (11:52):

I think in terms of the logistics of some of these products, cold chain is something that comes up over and over again in the healthcare system, which basically means, if you can imagine a remote health facility, they may not have the equipment and the storage capabilities to maintain products with short shelf lives or require advanced refrigeration in order to store. So they actually rely on an instant delivery service such as Zipline's to actually receive these products that need to be kept cold for a while.

EW (12:32):

Antivenom makes a whole lot of sense because you needed to have it for just the right snake, it doesn't really last that long, and it has to be kept in just the right environment. And so I can totally see. But how long does it take? If I was a doctor in Rwanda and I called Zipline and said, "I need this," and we already had a relationship. Now I'm not going to assume we're setting up the credit card right now, but how long does it take?

WH (13:07):

We have a system where we actually receive orders over WhatsApp. And so all these doctors can just send us a text message essentially saying, "Hey, we really need this product." There's actually two types of deliveries that we do. One is called resupply and one is called emergency. We encourage health facilities to place resupply orders so that we can send them products when we're not super backed up and they can get what they need for the week. But of course there's always emergency situation such as a patient who comes in with a snake bite, as you mentioned, and it would take less than an hour, on average 30 minutes, from the text message being sent to the package getting delivered at that health facility.

EW (13:58):

Where does this stuff come from? Do you work with larger hospitals to be the providers or does Zipline generate, manufacturer, suck blood? I don't know.

WH (14:14):

We work with the national center of blood transfusion. So they're actually a centralized warehouse in Rwanda and they store a lot of these blood and other medical supplies. We actually have a relationship with a lot of these centralized locations where medical products are typically stored. Before Zipline came in actually, if hospitals and health facilities needed products from these centers, they would send a car over to pick these up. Instead, we actually now have our Zipline distribution serve as these central hubs. So we basically place orders with these larger warehouse suppliers and they provide us with the medical products that we request every week.

WH (15:02):

But also in addition to that, there are incidents where we were unable to fly, say because of regulatory restrictions like late at night, things like that, health facilities have actually come to supplying warehouses to pick up blood and other medical supplies. In addition to that, there have recently been shortages in blood supply because of coronavirus and other reasons. People stopped donating blood quite as much. So Zipline has been doing a lot of community outreach and actually hosting blood drives to encourage people to donate blood. So I think Zipline over the years has really played a more and more active role in these national healthcare systems that we serve in.

EW (15:56):

One of the things that was interesting to me was that you have local operators, you're not doing everything from a centralized place in the world. The people on the ground in Rwanda operate fairly independently. How do you find the right people?

WH (16:17):

That's a really great question. We often say that the operators that we have in country are really overqualified. We have doctors serving in our fulfillment centers and we have engineers with electrical engineering degrees, mechanical engineering degrees working as flight operators. And honestly, they've been so instrumental in the success of our company and the success of our operations. They're at the front lines, they're the ones who are interacting with customers directly, they're the ones who are repairing our drones if anything goes wrong and catching failures before they actually happen, or they actually cause damage. So it's a really rigorous recruiting process to find these people. But I think after... Zipline has made a name in the countries that we serve, and so people get really excited at the opportunity to work with Zipline.

WH (17:17):

Operators play an extremely important role. As I mentioned, there's two different types of operators, there's fulfillment operators and there's flight operators. Fulfillment operators are the ones who are interacting with the hospitals and our customers, and they're managing the packages and packing them, whereas the flight operators are inserting the packages into the bodies of our drones, they're assembling the bodies. There's actually a pretty unique feature of our drones. We have what we call LRUs or line replaceable units. They actually come apart. So the body, the wings, and the battery are three separate components of the drone. So every time we actually bring it to the launcher, the operators are responsible for assembling the three parts for checking the software and the hardware are in line and nothing looks off nominal and they're ready to fly. In addition to that, they're also responsible for communicating with the air traffic control of whatever local equivalent of the FAA is in that country, and deconflicting the airspace manually.

EW (18:38):

You went to Rwanda and got the operator training. What was that like?

WH (18:45):

It was an amazing experience. Zipline, pre-pandemic, encouraged that all new employees of the company would go to Rwanda to see our operations live. And it's one thing to see our drones deliver on video, it's another to see it in person, but it's another to see it actually operational, saving customers' lives every day. And it was amazing watching these operators move so swiftly and they took their jobs so seriously. Every little aspect of safety, if you're standing behind the operator and they're about to launch the drone and you're not wearing safety goggles, they'll turn around and be like, "Hey, we can't launch it. You have to wear your safety glasses." So they are seriously well-trained.

WH (19:34):

I originally went with the intention of just observing the operations and getting to know some people in person, but I ended up falling in love with the process so much that I just went through all the courses and got trained to be a flight operator at the end. It was super fun taking all the exams and really understanding what the operators are going through firsthand and really learning all the pain points that I think as engineers we often overlook. We have our own goals and targets, but they're often not what really matters to operators. So I think having that experience of actually operating and getting to know what the product looks like to our users was really, really important to a lot of decisions that we make and how we prioritize our work.

EW (20:29):

How do you get sub-centimeter GPS? Do you need four antennas on this?

WH (20:36):

We have RTK on our GPSs, that's real time kinematics. We have a base station which has its own GPS, and we have another GPS on our drones. So we're actually sending corrections using the space station GPS, which we've surveyed its position very, very accurately. And it's sending that correction over to our drones as they come in for recovery.

EW (21:02):

One of the things that's always been a problem with drone companies is they come and they go, and they come and they go, and the drones suddenly are no longer available. But you guys have been around for a while and you've, what is it, 200,000 fulfillments?

WH (21:22):

Yeah, that's right.

EW (21:24):

What's the hardest part of operations at that scale?

WH (21:28):

It's definitely been challenging scaling up our operations over the past few years, but I think one really big advantage Zipline has as a company is that we're different than, say DJI who releases one product after probably months or years of testing and development and that's the product that all the customers see. If they want to iterate on that, then they have to go through another cycle and then release that product to customers again. At Zipline, we are managing the entire, I guess, stack in a way, from the design process to the actual service and the delivery experience, which means that we don't actually have to cut releases of our hardware and our software. We are constantly pushing updates at every given point in time. And if we want to replace, say a feature on the wing, say one of our sensors, we can just push that update and then send all those new updates overseas and software... For software it's even easier of course, because you can push those updates remotely.

WH (22:42):

So I think this has really enabled us to figure out what kind of problems that we see in the field and then work on solving those specific problems and pushing out those updates so that our aircraft becomes increasingly more reliable over time. And there are so many problems that we've fixed over the past few years and the problems that we see today and experience most frequently look very, very different than they did when we started off. And I think it's really a testament to how hard the engineering team works and the process of continuous improvement, basically.

CW (23:16):

You talked about launching recovery, but what does the flight look like? Is it mainly autonomous? Does a flight controller have direct influence over, say the controls of the aircraft or camera view, that kind of thing? Or is it more like just monitoring its location?

WH (23:38):

Yeah, it's completely autonomous. Actually, the role of operators are only to monitor its location and its telemetry and other health. There is a series of checks that are constantly running on our aircraft, we call them exceptions. So it'll send an exception if anything goes wrong and operators can monitor the status of the aircraft in the sky. But since a single operator is actually responsible for managing up to, say 20 aircraft in the sky at a single point in time, we don't actually expect them to do anything about it. And the aircraft, say if one of the air data sensors is broken, it'll turn around and fly back home without anyone prompting it to. And if there's a catastrophic failure, it will automatically and autonomously pull a parachute, which is its last line of defense. So if all else fails, it will pull a chute and safely glide to the ground.

EW (24:39):

How are you talking to the airplanes? What radio?

WH (24:44):

We have multiple forms of communication. One is radio, which is a 2.4 gigahertz radio that we only have access to within a kilometer of our base station. This is what we call our line of sight radio, and it's really important for things like recovery, especially, because we're sending those commands, that sub-centimeter level GPS location at 25 Hertz. So we really rely on this radio to get the location of that drone at really, really high speeds. The line of sight radio is also really important for drones to talk to each other. If there's two within the same airspace, it'll actually run through a process called deconfliction and one will go lower and one will go higher as to avoid them and prevent collision.

WH (25:46):

However, at farther locations and distances, we can't rely on our line of sight radio. So we use cell and we have 2G, 3G and 4G capabilities onboard our aircraft. So we rely on cell to send telemetry back to the ground station at about one Hertz. In addition to that, if cell also fails, especially in the locations that we serve, cell can be really sturdy. And so we also have satellite communications and with the satellite communications, we can actually reliably get information back from the aircraft. But the latencies for satellite communication are much, much higher than for cell.

EW (26:35):

Of course.

WH (26:38):

For sure. And if that fails, our drones are designed to fly out by themselves. They know their destination at any point in time. So even if we can't tell where it's going or what it's doing, it knows what to do.

EW (26:53):

Cool. You're giving up payload space for all of those backup radios.

WH (27:01):

Yeah.

EW (27:02):

How do you decide on the optimization for payload space versus backup technology?

WH (27:17):

Zipline has always been a very problem-oriented company. The technology that we design for, they have hard requirements. At the very least, we have to serve our customers and we have to be able to carry a reasonably-sized payload. Starting with that requirement, I guess there wasn't really much optimization to go from there because we had this 1.3 kilogram, now 1.8 kilogram payload requirement, that we thought was reasonable to serve our customers and to deliver what they need. Building from there, for every layer of communication and redundancy that we have, these were all driven by external requirements. So we needed to communicate at high frequencies at short distances, which informed our decision to include our 2.4 gigahertz radio. And then to be able to receive telemetry over longer distances, cell was the obvious solution to that. Our aircraft has actually evolved throughout the years based on all of these new requirements and things that we believe were really useful for our aircraft, either to increase the reliability on our end or to perhaps meet the demands of whatever regulatory body is monitoring and imposing their own requirements on our aircraft.

EW (28:57):

Regulatory bodies, yes. You operate in Rwanda, Ghana and the United States. Did I miss any?

WH (29:07):

Yes. That's right.

EW (29:10):

Which of those is the hardest?

WH (29:15):

In terms of regulatory bodies, the FAA definitely has the strictest and longest requirements. It's pretty interesting because the FAA is used to regulating manned aircraft. So as a lot of drone companies come into the space, we are actually helping to write these new regulations for autonomous aircraft for the future. There are a lot of hard requirements that our current aircraft, the one that's serving in Rwanda and Ghana, don't exactly meet up to the standards of, so we've been working with the FAA constantly, trying to get certifications, trying to make little baby steps towards gaining their trust and demonstrating that we do have a reliable aircraft and that we can meet the standards that they require.

EW (30:14):

Does it feel like too much? Does it feel like the FAA is just throwing up hoops for no reason?

WH (30:23):

At the end of the day, the FAA has the interest of people on the ground safety as much as possible. The hoops [crosstalk 00:30:35]

EW (30:35):

There was no way you could say anything about that, sorry. That was a terrible question. Actually, about your airframes, are you starting to sell them to other people since you are getting some traction with FAA and other regulatory bodies? Where can we get an airframe? Can we?

WH (30:59):

Yeah. As I mentioned before, Zipline's technology has been really reliant on the fact that we can move fast and we can make iterations so quickly. So the aircraft that you see today, look nothing like the aircraft... Maybe it looked like it, but on the inside, it didn't really look anything like the aircraft of a few months ago. And in a few months it's going to have new updates again. So because our aircraft is constantly evolving, we don't really have that strong cutoff point where we're ready to sell to customers. On top of that, as I mentioned before, we have this complicated launch and recovery system. So it's not exactly a system that can be easily sold to other people. And does require a lot of maintenance on the backend by our engineering team. So you can't exactly buy it right now in the market.

EW (31:57):

How long is a drone in service, and what triggers it's decommissioning? You said there were multiple pieces, do they all get decommissioned at the same time or which one lasts the longest?

WH (32:11):

Because our drones are composed of hundreds of little parts, in addition to those three main parts that I explained earlier, we decommission every component according to its own life cycle. The propellers are the ones that actually get the most wear and tear. And that has a very consistent cycle of being swapped out to put in fresh ones and to make sure that it can last through multiple deliveries. But as far as the other components go, we pretty much fly them until they have a serious failure. And then we'll swap them out. These aircraft are designed to withstand multiple years of use, and we are still flying hardware that we've made and built years ago. So they're quite resilient. And the purposeful design is actually to swap out little components so that we save a lot on money and resources.

EW (33:11):

Money, how can you make a profit in areas that you're trying to serve that may be otherwise-

CW (33:23):

Underfunded.

EW (33:24):

... underfunded? Thank you. That was a good word.

WH (33:28):

Actually, as of a few months ago, our operations in Ghana have become cashflow positive. So as folks of perhaps a more developed world, we often have a perception that these kinds of advanced technologies are really expensive and drone delivery would be impossible anywhere else because it simply would cost too much to afford. But that's not true at all. Like Zipline, we have a really scrappy company and we've built our technology not to look flashy, but to serve its purpose and exactly that, no more, no less. And I think that's really come to be an advantage for our companies since I think the fact that Ghana is cashflow positive, is an amazing example of how these efficiencies and our design and also the dedication of our operators and serving our customers has really been really effective.

WH (34:40):

We currently work with public national healthcare systems. So the governments are our primary customer and the government has more money than any individual hospital or health facility or person. And so we definitely depend on these larger governing bodies to provide the financial support for our system, but it's definitely worked in the countries that we serve, and there's been a real impact there.

EW (35:11):

Having talked to a listener in Ghana who was familiar with Zipline, it sounded like they're pretty excited to have Zipline be part of the medical process. Have you gotten to see that?

WH (35:30):

Sorry. What do you mean specifically?

EW (35:35):

You mentioned saving lives, but there's also this idea of building trust in the community so that when they see an aircraft, it's exciting, it's good. It's not scary.

WH (35:50):

That's a very interesting point. And I think people who've never seen Zipline before, there's been a negative light painted on drones, that they're heavily militarized or there's no way they can do any good. But I think seeing and experiencing is really believing, and that's especially true for Zipline. There's actually a pretty interesting story about that. When we were first serving in Rwanda, there was a mother who would see our drones fly overhead and she thought to herself, "Wow, these people must be crazy." She's watching the kids get really excited over these flying aircraft, and of course it's a really exciting thing to see, but as a mother, she was like, "This newfangled technology was insane."

WH (36:49):

Actually, she gave birth at a hospital and she had to undergo a C-section. And during her surgery, she started experiencing hemorrhaging. And so she was bleeding severely as a result of the surgery and the hospital that she was at didn't have enough blood to save her from that surgery. So the doctors placed an order for a resupply of blood from Zipline. And the Zipline drones came in, they did their job, they delivered the blood and the mother's life was saved. And since then she said, "I get what these drones are doing, they're saving lives and they saved my life." It's common to have skeptics for any new technology that comes in, but as soon as people realize what these drones are doing, I think they are very excited about it. And they realize that these are not military drones and they're here to save lives.

EW (37:56):

For you, what do you do at Zipline?

WH (38:01):

I've done a variety of things since I joined. I primarily work on our testing systems, especially our hardware and the loop testers. For anyone who doesn't know what a hardware in the loop test system is, it's basically a bit of avionics. So we have essentially our drone and our ground systems on a table sitting in the office. And we run all of our software through this hardware in the loop tester before it gets deployed to a real life flight so that if there are any errors or bugs in the code, we catch them before we actually take down a plane in real life. We describe it as a virtual reality for our drones, since, as far as the avionics on the table are concerned, it's flying in Rwanda or in the U.S., for all it's concerned.

EW (39:04):

And you mostly write software.

WH (39:10):

Yes, that's right.

EW (39:11):

But your background is in mechanical and aerospace engineering. How did you get into software?

WH (39:18):

That's a great question. Honestly, when I first came into college, I didn't really know what I wanted to do. I knew I loved building things with my hands, but I wanted to do more of a robotics major, I guess, but I could only choose one of the three between mechanical, electrical and computer science. So I went with mechanical because it was cool. I took some computer science classes along the way, so I didn't come into Zipline completely fresh, but it was definitely a learning curve because it was the first time I had written code in industry.

WH (39:56):

And I think writing code for a cool hobby, Arduino robot car is very, very different than writing flight critical software that actually gets flown around the world. So there was a really steep learning curve coming in but with the help and mentorship of my team and just the fact that they believed in me and they were there to help me along the way, I think it's been incredible opportunity and incredible growth. So I've definitely learned a lot since joining Zipline about software.

EW (40:34):

If for some reason Zipline no longer existed, which we don't hope that happens, but where would you look for a different job? Are you interested in the industry or the application or the code or the flying? What's the cool part for you?

WH (40:50):

I really like the interdisciplinary aspect. Although I do spend most of my time writing software, one thing I really love about Zipline is the fact that I have been able to CAD and I do often touch the hardware and I screw in screws. I still am really physical with the work that I do, but I also get the cool component of writing software and getting that instant gratification that mechanical designers doesn't have since there's really long lead times between your design process and when you actually get to see it come to life. So I really love that intersection. So if I had to choose a different job, I would definitely continue looking for things in the robotics space and things where I can get my hands dirty, but also improve and hone my coding skills.

EW (41:41):

Is it weird to think of a drone like a robot? I know it's autonomous, but robots have arms. I don't know. I suddenly have a very limited mental image of what a robot is and isn't.

WH (41:58):

I don't think it's weird at all. I think everyone has a slightly different definition of a robot. It's like how when you say what's an alien, everyone has the common conception of an alien, but an alien could look like anything. Robots come in all different shapes and sizes. And I think what makes a robot a robot is the fact that you code it, it has some autonomous element to it, and it has some actuation element to it. And that last part is not even necessarily true, but in the case of drones, it certainly is true. I mean, how else does it move? We are controlling its control services and its motors. I think having the ability to press a button and watch it just move on its own is a really fascinating thing and it's a unique property of robots.

EW (42:52):

I have a couple of listener questions, which some of them I have already asked, but still. Exploding Lemur wanted to know if your flight controller is entirely homebrew or based on an open design, like PixHawk.

WH (43:08):

Our flight controller is entirely homebrew. We have our own guidance navigation control team, and we have our own filters and things that really accurately determine our position at any point in time and what to do with that. So not based on PixHawk.

EW (43:28):

Laughlin asked about your telemetry and radio systems, which we've talked about. But the follow up question he had was, how do those choices scale with the number of vehicles, especially 2.4, if you've got an operator juggling 20, are there multiple operators in that same area?

WH (43:51):

We do have a few operators on the ground who are monitoring, but since operators are mainly there for monitoring the health of the aircraft, if nothing is going wrong according to the aircraft themselves, the operator... We basically have our own UI and it's an iPad app, and operators will pretty much just overlook them. And if there is an issue, it will get escalated on the UI immediately and then operators can monitor its health. We found that what we currently have actually scales quite well to hundreds of deliveries that we have per day around the world.

EW (44:33):

What are the future problems? What is Zipline looking at next?

WH (44:40):

As we talked about earlier, working with the FAA as the next target that the company is looking at. We really want to deliver in the U.S. and we've had a recent partnership with Walmart, which we're really excited about. We just want to expand the access of healthcare and other instant supplies that are really important for folks in the country and around the world of course. As far as the next big technology, our current aircraft actually doesn't have perception on it. It's flying around the world completely blind, in a way.

WH (45:20):

So we rely on operators to deconflict the airspace on their own. So if air traffic control or if they witness an incoming intruder aircraft, they would have to command the drone to either come home or to parachute-land in an emergency situation. But we are actively developing our perception stack to have our own aircraft to detect other aircraft in the sky. And that's definitely a big move for delivering to customers in the U.S. and for, in our own sake in Rwanda and Ghana, to have operators not be the ones to deconflict the airspace, but to have the aircraft do that on their own.

EW (46:03):

Walmart, are you doing medical from Walmart or is this where you start leaving the medical world?

WH (46:11):

Our initial contract with Walmart is with its medical and pharmaceutical division. We do plan on continuing that mission of increasing healthcare access to people around the world. But if our partnership with Walmart goes well, we would of course be open to delivering all kinds of small parcels that currently get delivered in a big truck that consumes gas. We want to get it to customers not same day, but same hour.

EW (46:50):

Where are you flying in the U.S. right now?

WH (46:53):

We are currently located in North Carolina and we're doing small, simple deliveries there. It's just for PPE in these pandemic times and really more than anything, it's there to be a demonstration to FAA that we can deliver reliably and to gain that trust and improve that relationship. But as far as our Walmart deal goes, we are actually planning on having our first distribution center with Walmart in Arkansas, where their headquarters are located.

EW (47:32):

I understand why you might need this in Ghana or Rwanda, even if the roads are good. Maybe I totally am ignorant and it's not, but I think the roads in North Carolina seemed fine. Why do you need a drone? You mentioned the trucks and they're big, but a drone has to go up and then over and then down and then up and then over and then down. But a truck goes house to house. Christopher's looking at me like I'm crazy. Do you want me to walk this back so that it's less crazy?

CW (48:10):

No. Keep going.

EW (48:15):

Let me say this right. Why are you going into areas where the infrastructure and roads are good enough? It makes sense to go in areas where the infrastructure may not be as good, but why? Is it-

CW (48:32):

Just let her answer the question.

EW (48:33):

Yes. Why?

WH (48:36):

That's a really good question. It's a question we get frequently. I think that's actually... In Zipline's early days, our company marketed itself as a solution to locations with poor road infrastructure. That still holds true today. We have a big Slack channel for our company and we share really interesting videos and photos. And one of the ones that always strikes me is a health facility in Ghana. It's, right now, the rainy season there, and this whole facility is literally an island. It's surrounded on all sides by water and people have to wade through the water in order to get to these health facilities. And they say, "Without Zipline, there's no way we could serve patients who are in need because trucks can't get to these locations."

WH (49:34):

And I think it's still such an incredible visual story for me as an engineer sitting in the U.S. where I can drive everywhere and just a really great reminder of the work Zipline is doing. But there's something that is always missing in this picture, which is, what is healthcare logistics actually, and why is it such a big challenge? There's a tradeoff when it comes to healthcare logistics. If you want to have health facilities and hospitals, always have the supplies needed to serve patients readily stocked at any point in time. That actually means with products with really short shelf lives, that there's tons of waste generated by these hospitals. If you think about it, you have to have blood and plasma for every single blood type. There's eight different blood types and they all have a certain expiration date.

WH (50:39):

Waste is something in the back of everyone's minds in America, it just gets thrown in the dumpster and then people don't really think about it, but it's wildly inefficient. And I think with Zipline, having a centralized distribution center with instant delivery can really help to reduce that level of waste. It's a great solution for anyone to adopt. The statistics basically say it all. Rwanda has significantly decreased its blood waste and increased its medical supply access to all of its people and its wastes numbers are so much better than the numbers we have in the U.S., so it's definitely a thing that can help people. There are folks who live in the U.S. who live in really remote and rural places. And in a time like coronavirus where contactless delivery had become more important than ever, I think people who live very far away and who don't necessarily want to drive in person and pick up products, they can really benefit from it.

WH (51:56):

On top of that, there's also the vaccine delivery logistics nightmare that people who are following the coronavirus news often talk about as well. People are saying that the vaccine that is getting developed will require something like negative 90 degree storage in order to maintain, and with something like truck delivery, that might not really be possible to deliver within that kind of temperatures and the kind of shelf life span that these vaccines have. So I really think drone delivery will be instrumental in these situations.

CW (52:43):

I think we tend to think of the United States as very rich and connected, and it is, but there's a lot of areas, like you said, rural areas and there's tribal areas in Appalachia, various regions that are very hard to access and don't have a lot of medical infrastructure. Then you have one clinic for hundreds of miles. And like you said, not stocked with things that they need to have on hand. I could totally see your primary mission being applicable almost anywhere.

EW (53:12):

I totally agree. Let's see. I have a question here that I think is appropriate and that is, is Zipline hiring?

WH (53:25):

Zipline is absolutely hiring. We are hiring more than we ever have before, in fact, including our Embedded team, we are actively looking for people to join our mission and help solve really, really interesting problems.

EW (53:43):

Are you looking for remote or Rwanda or Ghana or North Carolina or San Francisco? You're a pretty spread out company. Where are you looking for people to be hired?

WH (54:02):

We are actually looking for people in all of those facilities that you named. You can check out our website and our careers page for up-to-date information. The list is constantly changing and growing as we hire more people and have more positions open, so I think that's really the best source of truth for that information. Actually during coronavirus, a lot of people are working from home, but a few teammates and I have been going to the office quite regularly and we are building a fiscal product at the end of the day and we aren't stopping for anything. We deliver an essential resource to people around the world. So if you want to work in person, I think this is a really great opportunity to work on something awesome.

EW (55:01):

You have openings for an embedded system software engineer and an embedded system software lead. If you interview them, what are you going to ask? Any hints?

WH (55:13):

The Zipline interview is pretty fun in particular. We try not to ask questions that you would find on LeetCode or Cracking the Coding Interview. We really want to get a sense of how people operate on a day-to-day basis. So we'll often ask questions about, how would you solve this problem? And it's a problem that we had the other day. Just real world applications and being really safety critical-minded as well.

EW (55:51):

Whitney, it's been really great to talk to you. Do you have any last thoughts you'd like to leave us with, or maybe a story from the field?

WH (56:01):

I really think one of the most unique parts about Zipline is that we are solving real problems and there's so many companies out there that are doing things for the sake of being a startup or being entrepreneurial, doing something flashy, but I just love Zipline so much because you can really see the impact that we've had. And I think there's a really great story that I love that exemplifies this, which is when we were first launching in Ghana, we were all prepared. We had a launch date, a bunch of media was going to come over, the president was going to come over to witness this launch. We were all set up and ready to go and the engineering team was ready to support.

WH (56:57):

And a few days before that launch date, there was a truck driver who got in a very serious accident. And he was brought into the hospital and he was in dire need of medical products that the hospital couldn't actually supply. So the hospital called up our distribution center, which hadn't officially opened yet, and they said, "Hey, we know you're not open yet, but we really need these medical products. Could you deliver to us?" And that hospital actually happened to be one of the first hospitals that we were planning on serving anyways, so we said, "Why not?" And we actually sent them a delivery. So it's definitely not just a for show thing. We really have improved the healthcare systems in the places we serve. And I think that's just a really amazing thing about Zipline.

EW (57:59):

Our guest has been Whitney Huang, embedded software engineer at Zipline. You can find them at flyzipline.com, or see the link in the show notes.

CW (58:10):

Thanks Whitney.

WH (58:10):

Thank you so much.

EW (58:11):

Thank you too, Christopher, for producing and co-hosting. Thank you too Rene, Laughlin, Exploding Lemur, and Andrei for their questions. Thank you to our Patreon supporters for Whitney's mic, thank you for listening. You can always contact us at show@embedded.fm or hit the contact link on embedded@fm. Now a quote to leave you with from Langston Hughes, "Hold fast to dreams, for if dreams die, life is a broken-winged bird that cannot fly."

EW (58:43):

Embedded is an independently-produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California. If there are advertisements in the show, we did not put them there and do not receive money from them. At this time, our sponsors are Logical Elegance and listeners like you.