Intermode Builds the Robot for You
CEO Arnold Kadiu explains why delivery startups should leave the dirty work to him
Running a delivery robotics startup is hard work, as any exec from Serve to Starship can tell you. You’ve got to raise funds, find your customers, design the software, appease regulators, chase down wayward robots, and all that’s just on a good day. So why also burden yourself with having to manufacture the robots in house? Intermode’s Co-Founder and CEO Arnold Kadiu thinks that physical work is best left to firms like his, where he and his team of ex-Ford engineers are creating an assembly line for delivery robots of different sorts.
Jonah Bliss: Let’s start off with the basics, what is Intermode, and what do you do differently than others in the robotics industry?
Arnold Kadiu: Intermode is building robotics hardware infrastructure. We make a universal mobile robot called The Modal. What’s different is that we aren’t trying to own a specific robotic use case, instead, we have a series of modular components that attach to our automotive-grade platform so people can have turn-key mobile robots in a variety of use cases. Most of our robots are used as PDDs in last mile delivery. We full service lease the hardware which means that Intermode will take care of everything from maintenance to repairs, so that our customers can focus on their business. We give our customers the universal basics and then we have specific areas where they can customize to make the robot their own. Our belief is that there would be infinitely more opportunities for mobile robots to be deployed in our lives if the hardware was easy to use, accessible and scalable.
JB: Tell us a bit about you and your team’s background, and how that led you to start the company?
AK: We’re a team of ex-Ford engineers who worked in a variety of areas within the company: electric and autonomous vehicles, vehicle testing, mobility incubator, etc. Before that we built highway legal, solar power cars at the University of Michigan. Between us, we’ve built and tested quite a few electric vehicles.
I’ve always loved mobile robots. When I first started at Ford, I remember seeing the first mobile robots in the factories, (new engineers are usually required to work in a assembly plant for their first three weeks) and then seeing the news about early mobile robots outdoors and thought to myself, this is the future. A true first principles approach to changing how the world works, but I was a research engineer who designed architectures at the time. I helped design structures that held the sensors for Argo AI, but didn’t know much implementation side of computer vision, sensors fusion, autonomous navigation. I didn’t know how I could really contribute to the industry. Much to my frustration, these robots didn’t take off at the speed I thought they would, especially outdoors.
As I spent more time understanding how the auto industry worked, especially within commercial vehicles, I started to realize that there is a missing component for mobile robotics to grow: scalable, durable and affordable hardware. Mobile robotics needed its commercial vehicle, and I knew we could make it.
JB: What are the types of projects you’ve taken on?
AK: It’s been a surprising variety. Our first focus was to work on last-mile delivery robots where we provided the strongest value proposition, and that’s where the majority of the work has been. It’s still incredible to see a robot you’ve built teleoperated around the world at 15mph. The customers here had either built their own custom robot before and switched to ours or switched from a competitor. We’ve also had a few customers use our indoor variation in a factory setting. Another uses it as a mobile cell tower. My favorite, however, is a robotic crossing guard. Here’s a use case where there isn’t a big enough market to generate strong VC interest, but there’s a clear use case and need in the market that we enable by having a hardware foundation to build off of.
JB: How are you seeing hardware capabilities improve as the industry matures?
AK: I think the biggest change in PDDs will be focusing on hardware that can start doing more deliveries per hour with a wider range of customers and conditions, while still using the same vehicle. The small moving cooler design has won so far. It’s the cheapest to build, but what these operators care about is not the cheapest to build, it's how can I complete the most amount of profitable deliveries with it.
In order for these robots to be used in a wide range of use cases, they need to operate in all four seasons, carry American-sized amounts of groceries and be durable enough to handle all of the terrible infrastructure we see in the world. This robot starts looking more like a cargo ebike than a cooler.
Most importantly, it needs to move safely and quickly while doing so. A robot going 15 mph in the bike lane vs one operating 3 mph on the sidewalk is a widely different problem, but a robot going that fast can complete anywhere between 2-4x more deliveries a hour. Ali Ahmed has a great TechCrunch article on it.
In general mobile robotics, I think the greater problem is having the hardware fully integrated and ready to be deployed, instead of having to tinker with a variety of options.
JB: What do you believe are the advantages a robotics deploying startup has if they turn to a firm like yours, versus them doing it in house?
AK: Ultimately it comes down to, what do they want to be good at? We want to be good at making and taking care of hardware.
The biggest advantage initially to using a supplier is the ability to cut right to something that works well and easily grows with them, rather than divide their limited capital going through the hardware R&D process themselves. A startup can either run leaner without those hardware engineers or those engineers can instead be sales, computer vision, operations, etc.
Once they want to scale a different set of issues pop up: service + volume. Who’s going to take care of the robots in other markets, how many robots should I build, who makes spare parts, how many spare parts, etc. Again, instead of the costs to make that happen, they can grow their startup on demand by leveraging a supplier. UPS, FedEx, and more don’t make their own hardware, instead, they optimize their operations around hardware that best enables their business.
You can use a contract manufacturer, but the difficulty here is that if you’re making something truly bespoke, you don’t typically have the volume to make it cost-effectively. High-volume production is the only way to get around the old saying of, pick two: durable, effective, or cheap.
JB: The industry seems to be split in a big of a divide between teleoprations and autonomy; I feel like you might be neutral to this debate in the sense that you’ll build either for the right client. With that said, how do you feel about the pros and cons of the two models?
AK: There certainly is a split. The future state will always have some percentage of both, as laws in many places require human monitors, at least in the short term. So, I think the question most companies today start with is: do I need mostly autonomy to make my business viable?
Let's look at some simple numbers for making a teleoperated robot work. Let's say the robot's physical costs are around $850 a month all in. Let's take that robot, and if we can make an average of 15 deliveries a day, 7 days a week, that’s $1.90 per delivery. Now let's hire remote drivers from South America. That comes to roughly $2.50 per delivery. Data costs another $1.5 per delivery. Other various costs come out to roughly $0.5 per delivery. So all in, it’s a delivery cost of just under $6.5, with no tip required. These numbers are only improving as the cost of data continues to fall.
The thing I’m neglecting in the calculation above is speed. Your labor and data costs roughly stay the same, regardless of what speed you’re going. If your PDD only goes an average of 2 mph on its deliveries, that means the average distance to do this many deliveries in a day must be under ~0.35 miles away to make the math above work.
So if your robot can only go 2-3 mph, your only option to make it more profitable is to reduce the cost of labor, aka to automate it.
JB: You seem to focus on speed and getting robots off of the sidewalk, is there a particular reason why?
AK: I think sidewalks can be a hard place for mobile robots. The last 30 feet need to be on there, people don’t want to come out to the curb, but the infrastructure isn’t really designed for wheeled vehicles going more than a walking pace. The moment you go much faster than walking speeds, and it's crowded with people, it can get tricky and I think that’s one of the reasons you see backlash in some cities. If no one is around, I think sidewalks are fine, even if you’re going faster.
The biggest reason is I want robots to eventually replace ¼ of car trips in the US. That means less space taken up in our lives for parking lots, more space for people and businesses, far fewer cars on the road, and a dramatic drop in emissions. All while making our lives more convenient, it’s a win-win in my book.
You can’t do that if you’re traveling 3 mph and competing with pedestrians who want to walk places. Bike lanes and the edges of the road are much better as the robot can solve more real-world problems by operating in those conditions. There the traffic is more predictable, the infrastructure is (usually) better, and it’s easier to show a 1:1 benefit to cities to invest in more bike lanes.
JB: Are there other particular members of the community you find yourself working with a lot? Hardware or software vendors, regulators, prognosticators?
Keep reading with a 7-day free trial
Subscribe to OttOmate to keep reading this post and get 7 days of free access to the full post archives.