TechDrive Zurich

Collecting more robot data is not the same as learning from it.

Raphael Levy of Foxglove joins Justinas Mišeikis for an in-car conversation about data quality, Physical AI and the practical work of bringing robotics products to customers. He explains how his background in autonomous-driving simulation led him to data and Foxglove, and why engineering alone is not enough to build a successful robotics business.

In this episode:

• Turning robot data into a useful learning loop

• Moving from research to real customers

• Learning through pilots instead of waiting for perfection

• Finding markets beyond the obvious robotics use cases

• Bringing domain experts into technical teams

Watch TechDrive Zürich on YouTube: https://www.youtube.com/@TechDrive-Zurich

Explore the show and all listening platforms: https://zurich.techdrive.tv/

Join the WhatsApp community: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

A robot can do a backflip. That does not tell you whether it can load a dishwasher or solve a customer’s problem.

Andie Zhang joins Justinas Mišeikis for a drive through Zürich to discuss the practical work behind useful robots. Drawing on her experience at ABB Robotics and Tetra Pak, she explains why companies should start their first automation project with a small task, how workflows shape deployments, and what live demos can reveal.

We also discuss hiring beyond the usual robotics CV, testing products with unfamiliar users, and connecting product features to value a customer will pay for.

In this episode:

• Choosing the right first robot task
• Collaborative robots and the surrounding workflow
• Humanoid attention and responsible product claims
• The live demo that became a robot dance
• Job descriptions, diverse perspectives and user testing
• Selling customer value rather than a feature list

Join our WhatsApp group for new episodes and behind the scenes: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

ABOUT THE GUEST

Andie Zhang spent over 10 years at ABB Robotics and has released the first ABB Cobot YuMi. She's Global Portfolio and Technology Manager Collaborative Robots and has recently be recognised as one of top 100 Women in Robotics in Switzerland

ABOUT TECHDRIVE ZÜRICH

Candid, English-language conversations in a moving car with the founders, researchers, investors and operators shaping technology. Hosted by Justinas Mišeikis in Zürich, Switzerland.

WATCH, LISTEN AND FOLLOW

• Website: https://zurich.techdrive.tv/
• YouTube: https://www.youtube.com/@TechDrive-Zurich
• LinkedIn newsletter: https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7475433043554598912
• LinkedIn: https://www.linkedin.com/company/techdrivezurich/
• Spotify: https://open.spotify.com/show/033uReCOJSJ3CFbj8W5gDX
• Apple Podcasts: https://podcasts.apple.com/us/podcast/techdrive-zurich/id1896934782
• Amazon Music: https://music.amazon.com/podcasts/3168e5f7-4ba6-402d-81a0-0ec5b66282e0/techdrive-zurich
• Castbox: https://castbox.fm/channel/id7276552
• Pocket Casts: https://pca.st/2egd0gr2Instagram: https://www.instagram.com/techdrivezurich
• Facebook: https://www.facebook.com/profile.php?id=61591083008421
• TikTok: https://www.tiktok.com/@techdrive.zurich

Does moving AI data centres into space solve the problem, or change the engineering challenge?

Bilge Demirköz joins Justinas Mišeikis on TechDrive Zürich for a conversation about cooling, crowded orbits and the choices we make as AI becomes part of everyday life.

Bilge is a Professor of Physics at METU and a member of the UN AI Panel. Her perspective comes from working with AI in research while questioning where it is useful, what it consumes and how it affects people.

We discuss why solar energy does not make heat removal automatic in orbit, how satellite debris can create further collision risks, and why large infrastructure plans need a broader environmental assessment. Back on Earth, Bilge reflects on children's motivation to learn, chatbots that tell us what we want to hear, and the difficulty of building evidence while the technology keeps changing.

There is enthusiasm here too: analysing International Space Station data in the search for dark matter, an AI-assisted filter design she says worked despite looking unfamiliar, and student projects exploring the stratosphere. She also explains the local Qwen model in her lab and acknowledges the hallucinations the team still catches.

In this episode:

• Physics, mathematics, music and learning across disciplines
• Power versus cooling in orbital AI infrastructure
• Satellite collisions and shared environmental risks
• AI as fire: useful, powerful and in need of responsible use
• Education, curiosity and chatbot behaviour
• Scientific evidence and the UN AI Panel's role
• Her lab's local-model exception
• AI-assisted simulation and engineering design
• Stratospheric balloons and experiments closer to Earth
• The case for keeping an efficient solution that already works

Join the TechDrive Zürich WhatsApp group for new episodes and behind the scenes:

https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

ABOUT THE GUEST

Bilge Demirköz is a Professor of Physics at Middle East Technical University (METU) and a member of the UN's Independent International Scientific Panel on Artificial Intelligence. This episode shares her personal perspectives and research experiences.

ABOUT THE SHOW

TechDrive Zürich is an English-language series of candid, in-car conversations with founders, researchers, investors and operators shaping technology. Hosted by Justinas Mišeikis in Zürich, Switzerland.

Video episodes on YouTube: https://www.youtube.com/@TechDrive-Zurich

All episodes and platform links: https://zurich.techdrive.tv/

Follow on LinkedIn: https://www.linkedin.com/company/techdrivezurich/

Instagram: https://www.instagram.com/techdrivezurich

TikTok: https://www.tiktok.com/@techdrive.zurich

Follow TechDrive Zürich on your podcast app for the next drive.

How do you turn a robotics breakthrough into something customers actually use?

Max Vogt of Swisscom Ventures joins Justinas Mišeikis for a drive through Zürich to explore the gap between great research, a promising startup and technology that becomes part of everyday life.

Join the TechDrive Zürich WhatsApp community for new episodes and behind the scenes:

https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

Max’s route into venture capital began with law, took him to Hong Kong, and included helping students find their feet at St John’s College. That experience still shapes how he thinks about backing people.

We talk about what he looks for beyond a founder’s pitch deck, why technical teams need to stay flexible about their solution, and what seeing technology deployed in Shenzhen made him reconsider about Switzerland. We also discuss robotics and physical AI, the timing of a startup’s market, and why the outcome of a venture investment takes years to judge.

In this episode:

• From law and Hong Kong to venture capital
• Understanding the person behind the pitch
• Falling in love with the problem
• Robotics, physical AI and whether customers are ready
• Bringing Swiss technology into everyday use
• The work, relationships and patience behind a venture fund

TechDrive Zürich is a series of candid, in-car conversations with the people building technology in and around Zürich, hosted by Justinas Mišeikis.

Watch TechDrive Zürich on YouTube:

https://www.youtube.com/@TechDrive-Zurich

Connect with Max:

https://www.linkedin.com/in/maximilianvogt/

All episodes and listening platforms:

https://zurich.techdrive.tv/

LISTEN & FOLLOW TECHDRIVE ZÜRICH

• All episodes and platforms: https://zurich.techdrive.tv/
• Spotify: https://open.spotify.com/show/033uReCOJSJ3CFbj8W5gDX
• Apple Podcasts: https://podcasts.apple.com/us/podcast/techdrive-zurich/id1896934782
• LinkedIn: https://www.linkedin.com/company/techdrivezurich/
• Instagram: https://www.instagram.com/techdrivezurich
• TikTok: https://www.tiktok.com/@techdrive.zurich
• Facebook: https://www.facebook.com/profile.php?id=61591083008421
  

A robot can land a backflip, yet struggle to pull keys out of a pocket. Enzo Wälchli, founder of Operator Circle and former Chief Commercial Officer at ANYbotics, joins Justinas Mišeikis for a drive through Zürich to explain the gap between an impressive demo and a dependable product. From selling drilling machines at Hilti to building a robotics company’s commercial organisation, his experience shapes how he thinks about customers, hardware and investment. We discuss why deployments take time, what years of field data are worth, and where Switzerland has an opportunity in physical AI.

For first insights and behind the scenes, join our WhatsApp group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

In this episode:

• From a 23rd-floor strategy office to meeting customers on construction sites
• Why robotics companies need different expectations from SaaS businesses
• What industrial pilots reveal that laboratory demonstrations cannot
• Whether robot software can succeed without close hardware partnerships
• Humanoids, useful dexterity and the role of spectacular demos
• Switzerland’s engineering strengths and the thinking behind Operator Circle

TechDrive Zürich features candid, in-car conversations with the people building technology in and around Zürich, hosted by Justinas Mišeikis.

• Subscribe on LinkedIn: https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7475433043554598912
• Join our WhatsApp group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z
• Watch the video on YouTube: https://youtu.be/1ZWaSGfw--M
• Connect with Enzo: https://www.linkedin.com/in/ewaelchli/
• TechDrive Zürich website: https://techdrive.tv/

At a rooftop barbecue at CERN, a group of researchers asked each other what any of their work changed for ordinary people. One answer that came out of those evenings involved body bags, and the packaging used to protect particle detectors, and giving forensic teams enough time to return names to families.

Daniel Dobos trained as a particle physicist and spent most of his career at CERN building silicon and diamond detectors. The data rates forced him into machine learning long before the current wave: the ATLAS experiment produced 80 terabytes a second, far too much to store, so the analysis had to happen in real time. He went on to co-found THE PORT, a set of humanitarian hackathons run with challenges from the United Nations and the International Red Cross.

The technical heart of this conversation is a claim about understanding rather than cost. Daniel argues that we do not know how learning works, so we substitute scale for the missing knowledge, and that scale turns out to work surprisingly well. Good enough for fluent text, not good enough to call the problem solved. He points at reasoning benchmarks where a ten year old still beats the frontier models, and at the human brain running on 20 to 50 watts, as evidence that a far more efficient answer exists and we have not found it yet.

In this episode:

• Why 80 terabytes a second forced CERN into real time inference on FPGAs, not GPUs
• Forty million inferences a second, years before the AI hype
• The rooftop barbecue that moved him from pure research into the ecosystem
• THE PORT, and how detector packaging ended up giving families back their dead
• Why a hackathon team working on encryption recruited a magician
• The Harvard study suggesting one generation may not be smarter than the last
• What social media should have taught us about scaling a technology
• Why he let his children play with AI image generators before saying anything
• The question they arrived at on their own, and why that mattered more than a warning
• Fire as the oldest example of a technology that runs both ways
• Radiation in orbit, and why a flipped bit is worse for a transformer than for deterministic code
• The brain on 20 to 50 watts, and what neuromorphic and energy based approaches might change
• Learning overnight, and what dreaming could mean for a language model
• Fifteen minutes in a robotaxi in Shanghai, and how quickly an eleven year old stopped noticing

About the guest: Daniel Dobos is a particle physicist who spent most of his career at CERN working on silicon and diamond detectors and on real time machine learning for the ATLAS experiment. He is a co-founder of THE PORT humanitarian hackathons and a father of three.

About the show: TechDrive Zürich is a series of candid, in-car conversations with the people building technology in and around Zürich, hosted by Justinas Mišeikis.

🔗 Connect

• Guest: Daniel Dobos - https://www.linkedin.com/in/danieldobos/
• TechDrive Zürich, hosted by Justinas Mišeikis - https://techdrive.tv/
• WhatsApp Group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z
• Follow the LinkedIn newsletter for the whole series: https://www.linkedin.com/newsletters/techdrive-z%C3%BCrich-7475433043554598912/
• TechDrive Zurich website: https://techdrive.tv/
• YouTube channel: https://www.youtube.com/@TechDrive-Zurich

Cleaning robots are real. Greeter robots are real. Anything with an arm is not, at least not as a product you can buy and trust to work unsupervised.

Prof. Dr. Yulia Sandamirskaya said it precisely: nobody in the world has yet dared to leave a robot with a manipulator doing real work in a care facility and then gone to bed. That is the line between a demo and a product, and the industry has not crossed it.

She is a professor at ZHAW, the Zurich University of Applied Sciences, where she leads the Centre for Cognitive Computing in Life Sciences, and she is spinning a company out of her lab to bring robots into elderly care. So this is not scepticism from the sidelines.

We get into neuromorphic computing and why a fruit fly performs acrobatics on roughly 100,000 neurons and almost no power. Then the line that reframes everything: your eye is sharp across only about two degrees and jumps three times a second, so the steady scene in front of you is 95% imagination. Intelligence, she argues, is closer to how many things you can imagine at once than to how much data you collected.

She is blunt about AGI. It has been almost here since the Dartmouth summer school in 1956, and bigger static models do not close the gap. She makes the case against humanoids on legs, because a robot on stairs is a hazard to everyone near it, and describes a Neuraverse where robot skills work like apps.

Her team toured a care home, ended up in the staff room, and she asked a caregiving team leader almost as a joke why he needed robots at all. He paused, then started listing, and she had to get her notebook out. The people who will work beside the machine know where the real bottlenecks are.

She also explains why the business model has to be robot as a service rather than a purchase, and tells the story of her grandmother in Belarus, living alone and refusing help, which is the actual reason she is building this.

We just started a TechDrive Zürich WhatsApp group. Every episode the moment it lands, behind the scenes, the bloopers that did not make the cut, and updates straight from the guests themselves.

• WhatsApp Group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z
• Watch the full episode on YouTube: https://www.youtube.com/@TechDriveZurich
• And follow the LinkedIn newsletter for the whole series: https://www.linkedin.com/newsletters/techdrive-z%C3%BCrich-7475433043554598912/

• TechDrive Zurich website: https://techdrive.tv/

When Kateryna Portmann was a kid in Ukraine, she asked her parents for a robot every time they travelled. They could never find one, so she got dolls instead. She works in robotics now, as Senior Product Manager at ANYbotics, and she runs the Women in Robotics Society.

Her route there was not a straight line: international business, an executive MBA, consulting, a PhD in Hong Kong she could not finish, then a startup where a language barrier pushed her towards numbers and into product management. She landed in robotics in 2019, for a reason that never shows up on a CV. Her sister has cerebral palsy and used neurorehabilitation devices, which is how she found the company that made them.

The core of this conversation is what actually breaks a robot deployment. Assume the robot works perfectly. It has been simulated, tested and deployed correctly. Then people leave a ladder in the corridor, prop open the door that has to stay shut, and walk the stairs while the robot is on them. The machine is rarely the hard part. The building, the workflows and the people are.

In this episode:

• Why the robot is usually the easy part of a deployment
• Nobody wants another interface, and what that means for integration
• Robot ready, facility ready, organisation ready, and why all three have to line up
• The gap between three robots and a hundred, and why most pilots never cross it
• Why almost nobody has twenty years of mobile robotics experience, and why that helps
• Compliance running behind the technology, and building safe by design anyway
• Moving from brownfield sites today to greenfield in three to five years
• How to get into robotics now, and what these companies are actually hiring for
• Serving a population that is half women with teams that are twenty to thirty percent

Join exclusive WhatsApp Group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

You'll get sneak peek into upcoming guests, behind the scenes and more.

About the guest: Kateryna Portmann is Senior Product Manager at ANYbotics, working on robots for hazardous zones and critical infrastructure, and runs the Women in Robotics Society. She was previously at Hocoma in neurorehabilitation robotics.

About the show: TechDrive Zürich is a series of candid, in-car conversations with the people building technology in and around Zürich, hosted by Justinas Mišeikis.

▶ Watch the video version on YouTube: https://youtu.be/uvoba2PyXEQ

🔗 Guest: https://www.linkedin.com/in/katerynaportmann/

🔗 Women in Robotics Society: https://www.linkedin.com/company/women-in-robotics-switzerland

🔗 ANYbotics: https://www.anybotics.com/

🔗 TechDrive Zürich: https://techdrivezurich.lovable.app/

🔗 Hosted by Justinas Mišeikis: https://www.linkedin.com/in/justinasmiseikis/

For the first time on TechDrive Zürich there are two guests in the car!

Katka Letzing and Anna Hakobyan co-initiated Swiss {ai} Weeks, five weeks of AI events running across Switzerland from 1 September into early October, with more than 200 partners and more than 200 events behind it. We recorded two weeks before it opened.

It takes about twelve months to build, and they start the week after the previous one ends.

Most of the conversation is about restraint rather than acceleration. Anna explains why generating ten pages of a document in two minutes does not mean you should, and what happens when the person receiving it summarises it with AI. Katka explains why leading with AI stopped being impressive the moment everyone started doing it. Both of them describe the shift away from large frontier models toward smaller specialised ones, and the gap between organisations that pilot and organisations that scale.

They are also unusually direct about themselves. Katka says she handles negotiation and diplomacy because she was trained to negotiate in war zones, then says in the next breath that the technical side is not her strength. Anna talks about taking ego out of a project you care about, and admits that asking for help is the part she still works on.

And one number worth sitting with: partner organisations taught 6000 children about AI, and then those children went back to classrooms where the teachers had not had the same training.

In this episode:• Why they built a movement instead of a programme• Coming in with the mindset of an experiment• From Czechoslovakia and the UN to Zürich• Why organisations backed this with no direct business return• What people keep getting wrong about Apertus• Why smaller specialised models are winning• AI consciousness, and generating less• Whether hackathons actually go anywhere

⏱️ Chapters

00:00 Two guests, two weeks to go

01:44 It takes 12 months to build five weeks

02:54 Why a movement and not a programme

04:21 Come in with the mindset of an experiment

05:35 Katka: Czechoslovakia, Silicon Valley, Kickstart

07:46 Anna: the UN, Montreal, New York, tech

10:34 What each of them brings to the team

11:29 Taking ego out of the equation

13:29 Getting partners on board

14:32 They backed it with no direct business link

15:03 Why this works in Switzerland16:18 What would make this year a success

17:22 6000 kids, and the teachers who were not ready

18:41 Where AI is actually heading

19:42 The Apertus misunderstanding

20:51 Away from the frontier models

21:53 Ten pages in two minutes

22:55 AI does not make you cool any more

23:43 Do hackathons actually go anywhere

24:47 How to take part

Join exclusive WhatsApp Group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

You'll get sneak peek into upcoming guests, behind the scenes and more.

Swiss {ai} Weeks: https://ai-weeks.ch/

Katka Letzing: https://www.linkedin.com/in/katkaletzing/

Anna Hakobyan: https://www.linkedin.com/in/annahakobyan/

More of TechDrive Zürich: https://techdrivezurich.lovable.app/

Watch the video version on YouTube: https://youtu.be/mS4hiMQVwpk

TechDrive Zürich is hosted by Justinas Mišeikis:

https://jmiseikis.lovable.app/

You have probably seen the videos of a robot rallying against a real table tennis player. This episode is about the five years, and the team, behind it.

Join exclusive WhatsApp Group: https://chat.whatsapp.com/JgGidD8SqsuFzohXvZmo5z

You'll get sneak peek into upcoming guests, behind the scenes and more.

In this TechDrive Zürich drive, Mireille El Gheche shares a path that starts with electrical engineering in Lebanon, runs through a master's in radio communication in Paris, a PhD in optimisation, a postdoc in machine learning in Bordeaux and a stint at EPFL, and lands at Sony AI. The recruitment was unusual: nobody would tell her what the project was. They sent her two research papers, one on computer vision and one on table tennis from DeepMind, and let her connect the dots.

She guessed right, joined as a team lead on reinforcement learning, and ended up leading 10 engineers and researchers on a problem that had not been solved before.

The technical heart of the conversation is the gap between a system that works in simulation and a machine that has to survive the real world. Every component worked on its own. Making them work together, in real time, against a fast spinning ball, is where the five years went. Or as Mireille puts it: in the digital world you can reset and restart, but in the real world the physics does not forgive you, and if you are late by a few milliseconds the ball is already gone.

In this episode:

• Getting recruited for a secret project with two research papers and no details
• Why the 80/20 rule breaks down when nobody knows the solution yet
• Research team and engineering team, working in parallel for five years
• The real bottleneck: making every component work together on a real robot
• Why building a robot is fundamentally different from building an AI system
• How the sim-to-real gap narrowed over five years, and why it will never fully close
• What technical leadership actually demands: clarity, trust across disciplines, and making different voices heard

• Going from individual contributor to leading 10 people, and learning to let go
• Where robotics is heading, from care and industry to self-driving cars and drones
• Would she be the patient for a fully autonomous surgical robot?
• Women in Robotics, and why whoever builds the future shapes the future
• Studying engineering in Lebanon in a class that was almost 50/50

About the guest: Mireille El Gheche is a robotics and machine learning researcher who led the reinforcement learning team on Sony AI's table tennis robot project, and is an active participant in the Women in Robotics community.

About the show: TechDrive Zürich is a series of candid, in-car conversations with the people building technology in and around Zürich, hosted by Justinas Mišeikis.