Documented project
PREEN Industrial AI, Robotics & IP Platform
PREEN Technologies · Nov 2020 to Nov 2022
Zero-to-scale CTO leadership spanning the patented core of a compressed-air droplet wash/dry, multi-camera vision, path-planning, and robotic apparatus, plus its 40-camera real-time 3D and 14-DoF dual-robot implementation, certification, team growth, fundraising, and IP.
Leadership mandate
- →Scaled the multidisciplinary technology organisation from 4 to more than 25 specialists.
- →As CTO, replaced a physically unscalable steam approach with a compressed-air droplet architecture designed to wash and immediately dry vehicle surfaces.
- →Owned technology strategy across robotics hardware, software, AI vision, product delivery, certification, and IP.
- →Connected hands-on engineering decisions with fundraising, commercialisation, and transaction readiness.
Hands-on technology scope
- →Specified firmware and Redis live streaming for 40 Intel RealSense cameras and an on-the-fly 3D reconstruction pipeline producing a vehicle object in about 30 seconds at 5 mm accuracy.
- →Directed ML-based semantic segmentation of the generated model to identify glass, doors, and other vehicle-surface classes needed for treatment planning.
- →Implemented real-time path planning for two industrial robots on linear tracks with 14 total degrees of freedom, moving them across the vehicle for the complete wash-and-dry sequence. That 3D generation and path planning was then the fastest in the world: nobody else was technologically able to turn 40 camera images into a 3D object at 5 mm accuracy and drive two robots in real time over a car.
- →Directed hardware/software integration, industrialisation, testing, and TÜV/CE-related certification work.
- →Developed and protected core invention concepts through coordinated hardware and software patent work.
Selected outcomes
Published patent families
His published patent families WO2022263238A1 and WO2022263239A1 publicly name Dr. Andree Welker as inventor and document the core inventions on cleaning method, vision, path planning, nozzles, drying, and robotic apparatus.
Public-source evidence
Organisation scale
Scaled the multidisciplinary technology organisation from 4 to more than 25 specialists.
Named written references
Commercial and funding scope
Supported 15× valuation growth and raised more than $20M in grants and more than $10M from VC/PE investors. Dated named written reference letters corroborate these figures; the letters remain private and can be provided for legitimate due diligence.
Named written references
Wash-and-dry architecture
The published patents describe pressurized gas atomizing water into accelerated droplets and compressed-air blow drying before, during, or after cleaning.
Public-source evidence
Patented 3D vision and path planning
The patent record describes an 8, 64-camera vision module, optical/infrared or stereo cameras, point-cloud and mesh generation, separation into individual surfaces, neural-network analysis, and cleaning-path calculation.
Public-source evidence
Implemented real-time vision pipeline
Integrated 40 Intel RealSense cameras, firmware, Redis streaming, 3D reconstruction, and semantic segmentation to generate and classify a vehicle object in about 30 seconds at 5 mm accuracy.
First-party operating data
Patented robotic apparatus
The patent record describes robot-arm-controlled nozzles, calculated cleaning paths, 7-axis actuation, rails, collision avoidance, and apparatuses with more than one robot arm.
Public-source evidence
Implemented 14-DoF robotic execution
Planned and executed real-time trajectories for two industrial robots on linear tracks with 14 total degrees of freedom for vehicle washing and drying. That 3D generation and path planning was then the fastest in the world: nobody else was technologically able to turn 40 camera images into a 3D object at 5 mm accuracy and drive two robots in real time over a car.
First-party operating data
Named commercial licenses
Licensed IP to Microsoft, Google, Intel, NVIDIA, AWS, and Fraunhofer. Dated named written reference letters corroborate these named licenses; the letters remain private and can be provided for legitimate due diligence. They are not independently audited public license registers and do not appear in the patent record.
Named written references
Relevant executive mandates
Sources & evidence
- 1.AI vision and cleaning-method patent family: Method for Cleaning a Surface
Espacenet / World Intellectual Property Organization · official record
His published patent family WO2022263239A1 publicly names Dr. Andree Welker as inventor and describes gas-atomized water droplets, nozzle design, robotic application, and compressed-air blow drying.
- 2.Robotics hardware patent family: Method and Apparatus for Cleaning a Vehicle Surface
Espacenet / World Intellectual Property Organization · official record
His published patent family WO2022263238A1 publicly names Dr. Andree Welker as inventor and describes multi-camera vision, point clouds and meshes, surface separation, neural-network analysis, cleaning-path calculation, 7-axis robot actuation, rails, atomized-liquid cleaning, and blow drying.
- 3.Named written references for scale and funding: Named written references, company-building scale figures
Private written references, available for due diligence · named written reference
Dated named written reference letters corroborate 15× valuation growth from CHF 7M to CHF 45M, CHF 40M+ fundraising exposure, teams scaled from 4 to 40+ specialists, and growth supported to CHF 50M revenue. The letters remain private and are not published; they can be provided for legitimate due diligence, subject to appropriate privacy handling. They are not independently audited public accounts.
- 4.Named written references for commercial licenses: Named written references, commercial IP licenses
Private written references, available for due diligence · named written reference
Dated named written reference letters corroborate named commercial licenses to Microsoft, Google, Intel, NVIDIA, AWS, and Fraunhofer. The letters remain private and are not published; they can be provided for legitimate due diligence, subject to appropriate privacy handling. They are not independently audited public license registers and must not be described as appearing in the patent record.
- 5.Implementation record: PREEN executive operating record
Dr. Andree Welker · first-party disclosure
First-party executive operating record covering compressed-air droplet wash/dry architecture, 40-camera real-time 3D reconstruction, semantic segmentation, 14-DoF dual-robot execution, and implementation figures not stated in the patent text.