
Kyungpyo Heo CTO
From a single drink to humanoids — drawing the five-year technology roadmap
From a single drink to humanoids — drawing the five-year technology roadmap
“What we're building right now isn't a beverage automation robot. It's creating, for the first time in Korea, a brand-new technology category called cloud robotics.”

The person who draws Li.fizz's technical vision and roadmap. Chipsets, actuators, the cloud robotics system, and humanoids — this is the seat that decides how deep the company will internalize technology.
Q. Could you introduce yourself?
Hello, I'm Kyungpyo Heo, CTO of Li.fizz. I completed my undergraduate degree at Hanyang University's Department of Convergence Electronic Engineering and my master's at KAIST's School of Electrical Engineering. The work I built during my Hanyang years won a CES 2024 Innovation Award. My core background is circuit and firmware design, but I've also worked on web development and data analysis. So I was naturally drawn to an environment like ours — one where every layer goes into a single product.
At Li.fizz, I'm responsible for the technical vision and roadmap of the entire company. It's the seat where you decide what to build, when, and on which technology stack.
Q. What does a CTO do at Li.fizz?
The CTO is responsible for the company's technical vision and roadmap. It's not day-to-day implementation — it's updating, every day, the answer to "what technology should we be standing on three or five years from now?"
Specifically, four things.
First, designing the technology roadmap. Deciding which products to build, when, and on what tech stack. We have a clear roadmap that expands stepwise from the bartender robot to cooking robots, and then to humanoids. Decisions like which technology becomes core at each stage, and when we launch what, all start here.
Second, deciding what to build in-house. We build everything — from chipsets to actuators to the cloud system. But we can't literally build all of it. "This has to be in-house to create a moat" vs. "this is fine to buy" — these judgments shape the next five years of the company.
Third, setting the direction of R&D. Our in-house chipset (the L1 Chip), in-house actuators (the PCB Stator Motor), the cloud robotics system, and the next-generation humanoid — I draw the direction each R&D track should take.
Fourth, external technical partnerships and academic collaboration. Payment data integration with Toss Place, advisory relationships with external experts like Professor Dongjin Lee, joint research with academia — all of this falls into this area.
In one line: it's the seat that builds "tomorrow's technology" in advance, so the company isn't trapped by "today's problems."
Q. What brought you to Li.fizz?
CEO Jungsoo Song and I go back to our undergraduate days at Hanyang's Convergence Electronic Engineering. After graduation, I went to KAIST for my master's, and Jungsoo built product-development experience through InBody, LuaLab, and Rabino. Both of us are circuit/firmware people, so we always talked in the same tone.
One day Jungsoo said, "I want to really solve kitchen automation." Looking at a market where existing robots cost 80 million won, took 60 seconds, and reached only 0.2% adoption, he asked, "This is clearly a problem worth solving — why has no one solved it properly?" I wanted to work through that question with him.
Especially because both of us are circuit/firmware people, we knew: this isn't simply about making a big robot smaller. It's a problem that resolves only when you redesign the chipset, the actuator, and the cloud — all of it. So we started from our own chipset, the L1 Chip.
Q. Could you walk us through Li.fizz's technology roadmap at the big-picture level?
Three stages, roughly. I describe it similarly in IR.
Stage 1: Bartender robot (done, currently in production). Deployed across 1,400 venues — running every day at the Pulmuone SkyHub Lounge at Incheon Airport, premium resorts and golf courses, Q-Q Sushi, Nuguna Holdak Banhan Dak, Norang Chicken, and more. This was the first validation stage for our cloud robotics system. It started with a single drink.
Stage 2: Cooking robot (development completed in 2025). The stage where we expand from drinks into food. Starting with a one-pack snack F&B cooking robot, we extend the lineup into frying and other cooking categories. From here, the core isn't a single robot — it's a structure where multiple robots collaborate through the cloud. When an order comes in, the bartender robot and the cooking robot start working simultaneously, and a load-balancing engine optimally allocates every resource in the venue. This is what we mean by real "cloud robotics."
Stage 3: A humanoid for kitchen automation. Not a robot that performs a single task, but one that flexibly performs many tasks like a human. This is the real game-changer. If today you'd need to put 10 robots into one kitchen, a single humanoid can replace them all. Our company's ultimate vision — over 90% kitchen automation — is completed here.
And from 2028, we begin global expansion into Southeast Asia and the US, carrying the cloud robotics system we validated in Korea.
Q. Building your own chipset (the L1 Chip) and your own actuator (the PCB Stator Motor) is a really unusual decision compared to most companies. Why go that deep?
I get this question often. "Can't you just buy off-the-shelf parts?"
The answer is clear. You can't deliver the price point and performance we promised that way.
There's a reason existing robots cost 80 million won. Industrial controllers, industrial motors, and the separate control systems they require — all expensive. We had to bring that down to 4.5 million won. A 17.7× gap. That became possible thanks to the L1 Chip — a controller we designed in-house, optimized for beverage manufacturing. We stripped out everything a general SBC carries that we don't need, and packed in only what our robot actually requires, hard.
The actuator is the same reason. We're developing our own actuator — the PCB Stator Motor. A general motor is bulky and expensive; in PCB form it's flat, thin, and cheap. The result is that our bartender robot fits into 900× less volume than competitors.
One step further: these in-house components carry directly into our next-stage products (cooking robots, humanoids). Core technology built once becomes the foundation for every product line. This is what we call vertical integration, and it's the part competitors find hardest to copy.
Q. Could you say more about the cloud robotics system?
This is actually the true core of our company. A single robot running smoothly in a store is, in a sense, only step one. The real problem we're solving is getting all the robots across 1,400 stores — and soon tens of thousands — connected through the cloud and moving like a single body.
Specifically, three axes.
Automation robot lineup — bartender, cooking, humanoid. Each robot collaborates through the cloud, gets remote feature improvements via OTA, and cloud AI learns the venue's situation.
Real-time order data integration — through partnerships with payment services like Toss Place, the robot starts cooking the moment an order is placed. Order load-balancing across multiple robots inside a venue happens automatically.
Store operations data analysis — we gather big data per venue and deliver insights to owners, like AI auto-ordering and store-tailored robot recommendations. Data learned in one store improves operations in another — that kind of network effect emerges.
We aim to land the first version of this system in 2025, with a proper AI layer arriving in 2026. The key point is that all of this runs on top of our in-house chipset and our in-house actuators.
Q. As CTO, what's the hardest decision you face?
**"What *not* to build."** Building, honestly, is easy. Our team is strong enough that if we set our minds to it, we can build almost anything. The really hard part is deciding "we won't build this — not now."
A startup's resources are finite. If you build A, you can't build B. So every week I ask, "If we build this now, what will it mean for us a year from now?" Some technologies, built today, become a moat three years from now. Others, even if built today, become meaningless within a year. Distinguishing between the two is the CTO's real job.
For example, starting humanoid R&D right now was a big decision. It doesn't help today's revenue. But if the market is heading toward humanoids in five years, not starting now means we'll never catch up. There's no right answer to a decision like this. CEO Song and I debate it every week.
Q. Who would you like to work with?
Three things.
People who look one step deeper into the technology. We're a company that builds chipsets and actuators. The stance "use the given library to ship fast" matters less here than "understand *why* that library is designed the way it is, and build a better one if needed." I especially recommend us to people from academia, and to senior engineers.
People who can move between vision and execution. Some people are great at painting a vision but weak on execution; others execute fast but their vision is narrow. The CTO line needs both fields of view. People who built something end to end during their undergrad years tend to be strong here.
People who think on a long-term horizon. We don't make decisions based on quarterly KPIs alone. Some decisions only make sense looking three years out. People who can sketch what an expensive R&D bet today becomes as a moat three years from now fit well.
Q. As CTO, what kind of engineer do you think will enjoy Li.fizz the most?
The kind who says, "I want to experience varied technical depth inside a single company." Companies usually specialize. Firmware engineers stay in firmware for life; backend engineers stay in backend. Li.fizz is the opposite. From chipsets to the cloud, from circuits to AI — you can touch every layer inside one company.
And our in-house AI-Agent-based development infrastructure means a single person can deep-dive into many areas quickly. That's our real differentiator. You can grow not as "an engineer replaced by AI," but as "an engineer who uses AI to work deeper and broader."
Q. Finally, a word to those considering the CTO line.
While finishing my master's at KAIST, I looked hard for "a company in Korea that does truly deep R&D." Honestly, there weren't many. So we're building that company ourselves.
What we're building right now isn't simply "a beverage automation robot." It's creating, for the first time in Korea, a brand-new technology category called cloud robotics. A Korean startup that owns its chipset, its actuators, its cloud, and its AI in-house is genuinely uncommon.
Building all of this from zero is an experience you can't easily get anywhere else. If you want to do deep R&D, or if you want to shape with your own hands what Korean industry will look like five years from now, I'd really like to have a conversation.
