Philipp Kaindl

I'm an applied scientist focused and interested in the parts of AI that still offer a lot of exploration: semantic data layers that give agents a grounding in the real world, multi-agent architectures that go beyond simple orchestration, and reinforcement learning — including vision RL — as a path toward models that genuinely learn from interaction. My background is in Mechanical Engineering (TU Munich), though I came to it via an IT education in Austria. I reason from first principles where I can, and the Feynman approach to learning and understanding shapes how I work more than any specific tool or framework — understanding something well enough to rebuild it from scratch rather than just trusting the abstraction.

Before AI I worked as an engineering consultant at Airbus APWORKS — an additive manufacturing startup spun out of Airbus — designing structural parts using CAD, CFD, and FEA. That mix of hardware constraints and computational tools got me hooked on the intersection of physical and software systems. I still follow 3D printing closely, especially as AI starts to reshape generative design.

In 2019 I co-founded Elixon, an EU EXIST and NVIDIA-funded startup doing real-time traffic optimization with LiDAR/camera sensor fusion. After that I moved through senior data science roles at UniCredit and Allianz before joining AWS — first as a Senior Solutions Architect for AI/ML across EMEA, then as a Senior Applied Scientist at the GenAI Innovation Center, building production AI systems with enterprise customers. To all my mentors I had during the times, a big thank you, in case you are reading this.

If you are looking for a hyper polished blog, this is probably not for you. I spend a lot of time typing and dictating my blogs, with some editing with the help of AI, but I try to keep it minimal.

I write here when I think I actually know something — not just when I've read about something. You'll find technical deep-dives, but also opinions and the occasional rant. You won't find the latest AI news here. There is twitter / X and LinkedIn for that.