Product for technology that touches the physical world.
I'm a robotics engineer who became a product manager because I kept wanting to own the whole
problem β the hardware, the software, and the decision about what's worth building.
Nine years across control systems, clinical robotics, and B2B platforms; currently shaping
the integration ecosystem at Frontify.
IROS 2022 published robotics researchEPFL medical robot, 0 β clinical pilots2Γ ARR, twice as first PM at Timlyβ100 integrations one ecosystem at Frontify
Fig. 01 β Live systemBall-balancing robot, planar model Β· state-feedback control @ 240 Hz running
ΞΈ lean 0.0Β°
αΊ ball 0.00 m/s
Ο motor 0.0 NΒ·m
balanced for 0.0 s
wind 0 N
Crashed. Even feedback linearization has limits.
A planar cousin of my master thesis β Control of a Ball-Balancing Robot
(Aalborg University, 2020): five degrees of freedom, three actuators, an underactuated MIMO system
modeled with quaternion dynamics, estimated by unscented Kalman filters, and stabilized by nonlinear
feedback linearization. This demo runs a 2-DOF simplification with LQR-style state feedback.
Click the canvas or use β/β to shove it. Add wind and notice the steady-state lean β this
controller has no integral action. The full treatment (and why the humble EKF beat my fancy UKF) is in the thesis.
Field reports
Ambiguous technical problem, shipped product. Four times.
EPFL Β· CHILI Lab Β· 2021β2023 Β· Lead Engineer & Product Manager
The robot that changed jobs
I was hired to add features to a tabletop educational robot and explore a spinout. Discovery with
clinicians surfaced something better: stroke-rehabilitation therapy. We pivoted β building a gamified,
tangible Pac-Man-like therapy on the robot's haptic and modular capabilities, validated with hospital
partners. To get there I led the full rebuild: mechanical drive redesign, new electronics, scalable
firmware, ROS integration, a Unity SDK for therapy games, and clinician-facing monitoring UIs.
In robotics, discovering the right use case is the product decision.
I inherited a hundred-odd integrations, fifteen-plus partners, and no operating model. I built the
governance, portfolio structure, and delivery function; redirected Frontify's first e-commerce
integration to a synchronization architecture after discovery broke the original concept; and
started the ecosystem's first usage-analytics foundation. Also: AI-governance discovery β content
credentials (C2PA), provenance, EU AI Act readiness.
β100 integrations, one model15+ partner engagements1st e-commerce integration
Timly Β· 2023β2025 Β· First Product Manager
Feature factory β product-led
First PM at a Zurich SaaS startup. I built the product function from zero β discovery cadence,
prioritization, Mixpanel/NPS instrumentation, session recordings β and moved the company from
shipping features to shipping outcomes. The numbers followed.
ARR & customers 2Γ YoY, twiceNPS 4.0 β 7.2
Fremco Β· 2018β2019 Β· R&D Engineer
Robots meet mud
Sensing and data-acquisition systems for fiber-optic installation machinery, tested in the field with
the operators who used them. Industrial reality is the best teacher of failure modes β lab prototypes
don't get rained on.
embedded sensing β production designfield-tested with operators
Selected systems
Things I've modeled, built, or published.
MSc thesis Β· Aalborg Β· 2020
Control of a Ball-Balancing Robot
Quaternion Lagrangian dynamics for a 5-DOF underactuated MIMO ballbot; unit-norm-preserving
unscented Kalman filters; cascade nonlinear feedback-linearization control benchmarked against
sliding mode. Honest finding: the simpler EKF won on complexity-vs-benefit.
The peer-reviewed account of the robot platform I led at EPFL β modularity, tangibility, and the
engineering that let one small robot serve classrooms and clinics alike.
Led business planning and market research for an autonomous painting concept in a startup incubator.
Revived in 2026 as wallbot β an open-source cable-driven wall robot with a flat-coverage mode
and a high-resolution mural mode.
A low-cost underactuated robotic gripper; ultrasonic fault detection for concrete and metals;
modeling and control of a hydraulic press. Project-based learning, Danish style.
mechatronicsNDTcontrol
β Current focus
Building wallbot β
a cable-driven wall-painting robot, hardware on the way β and
rebuilding the ballbot
in simulation to compare classical control against learned policies, while working through the VLA stack.
Notes will land here.
About
I started coding at fourteen because logic mapped cleanly onto outcomes. I studied software
engineering, then went back for a second degree in industrial engineering β because I wanted to
understand the physical world I was writing code for. Choosing years of a second foundation over
a faster paycheck is the most honest signal of how I'm wired: depth and optionality over the fast path.
Since then I've oscillated between deep tech β control systems, sensing, robotics, computer vision β
and the strategic side: consulting at Accenture and EY, then product leadership in Swiss SaaS.
At EPFL the two careers fused. I'm most useful at the seam where software, hardware, and operations
meet, translating hard technical reality into business decisions and back.
What I'm optimizing for now: a return to deep tech β robotics, AI, computer vision β in domains where
the work touches the real world: medical devices, construction, agriculture. The long-term goal hasn't
changed since I was eighteen: build my own products.
Based
Zurich, Switzerland Β· EU citizen Β· Permit B
Education
MSc Mechatronics & Control, Aalborg Β· BSc Industrial Eng., Carlos III Β· BSc Software Eng., King Juan Carlos
Languages
Spanish (native) Β· English (C1) Β· German (A2) Β· French (A2)
Mentoring
ADPList Β· Lenny's community Β· 8 mentees across product & engineering