Robot maker · early

General Trajectory

Corporate marketmanufacturinglogisticsCurated record

generaltrajectory.com

General Trajectory develops foundation models and reasoning models for industrial robot manipulation, fine-tuning vision-language models to output low-level robot controls for tasks including mixed-SKU palletization, picking, sorting, and packing. The system applies chain-of-thought reasoning before executing physical actions, enabling robots to handle complex multi-step logistics work without manual re-programming. Training uses reinforcement learning from verifiable rewards inside NVIDIA Isaac Sim, eliminating the need for human annotation. During YC W25 the team signed commercial LOIs with industrial robot operators and released an open-source low-cost teleop stack.