Banseog note · 2026-10-26
Building the 'brain' software for humanoids
Current stage · from public events · 2026-11-14
- R&D · prototype
- Pilot · first production
- Scaling production
- Operations · field rollout
- Going global · localizing
- M&A · integration
- Restructuring
R&D · prototype Long-duration autonomous missions demonstrated with Reflect v1.0, without commercial-customer deployment figures
- 2026-06-29 It unveiled Reflect v1.0 demonstrations of autonomous missions involving movement through multistory buildings, opening boxes and organizing objects. Source
Roles often seen at this stage Robot software, AI & vision · Controls & embedded · Mechanical & hardware design
Companies at the same stage → · Same stage in other industries →
- A humanoid autonomy software company founded in Zurich, Switzerland, in 2024. Using reinforcement learning and simulation-to-reality transfer (sim-to-real), it is building a hardware-agnostic autonomy stack that combines humanoid locomotion, manipulation and task decision-making.
- In November 2025 it raised a USD 50 million Series A with participation from DST Global Partners, NVentures (NVIDIA's venture arm), redalpine, Prosus Ventures and Moonfire; earlier the same year it had raised a USD 7.35 million seed round.
- With Reflect v1.0, released in June 2026, it reported a 16-step mission completion rate of 90% after reinforcement learning fine-tuning (38% with supervised learning alone), and said the robot climbed up and down stairs more than 100 times in a row.
Decisions within a year
- Expanding locations in November 2025 it said it would use the Series A to grow its Zurich R&D team, add compute and robots, and establish a US presence. Its website currently lists a Zurich head office and a San Francisco office in the US.
- Product on June 29, 2026, it released Reflect v1.0. With a single voice command, a humanoid carries out a long mission, using elevators and stairs to move between floors, find and open boxes, and put items away.
- Collaboration in July 2026, together with Niantic Spatial and NVIDIA, it announced a method that recreates a site in simulation from a single 360-degree camera capture and transfers a camera-based locomotion policy trained there directly to a real office. The success rate in the recreated environment of its own office was 97.8%.
People they will look for
- AI research engineer (robot manipulation, human data) — expanding Reflect's hand manipulation abilities and learning from human demonstration data
- Forward deployed engineer deploying and tuning the autonomy stack with partner robot makers and at customer sites
- ML and infrastructure software engineer — running growing compute resources and large-scale simulation training environments
For hiring managers
All published roles are based in Zurich, so the target is talent able to move abroad rather than hiring in Korea. In Korea, the robotics affiliates of large companies and humanoid startups are looking for the same reinforcement learning and sim-to-real researchers, so people who combine publications with real-robot experience should be approached early.
For job seekers
Experience transferring reinforcement learning policies from simulation to real hardware on humanoid or quadruped robots, and hands-on experience bringing up robot software at customer sites, are worth the most at this company right now.
An analysis written from public sources. Sources: Flexion website (mission, Zurich head office, San Francisco office), The Robot Report 2025-11-26 (USD 50 million Series A, investors, USD 7.35 million seed, founded 2024, planned use of funds), Flexion news 2026-06-29 (Reflect v1.0 figures), Flexion news 2026-07-20 (collaboration with Niantic Spatial and NVIDIA), Flexion careers page 2026-10 (roles, locations).
Open jobs
Banseog checks official company careers pages every day. A job is marked closed only when the official page returns a removal (404 or 410).