Forenly AI

I build the "skill layer" for humanoid - turning machines into systems that actually do work in the real world. Locomotion, manipulation, and whole-body control, trained in simulation and deployed to hardware like the Unitree G1. Background: 5+ years in aerospace engineering at Turkish Aerospace (TUSAŞ) + Mechanical Engineering at ITU. I understand physics and hardware before writing policy code—exactly what Physical AI demands. What I deliver: 🤖 Humanoid policies — locomotion, reaching & manipulation as deployable ONNX brains (MuJoCo, LeRobot) 🌍 Scene-to-Sim — real-world captures → robot-trainable simulations via Gaussian splatting + collision meshes 🧠 Imitation & RL — data-efficient training (LeRobot ACT/Diffusion/SmolVLA), teleop data, sim2real transfer Working on humanoids, embodied AI, or anything where software meets the physical world? Let's talk.