#01
RoboPick
Voice-controlled pick-and-place on a UR5e arm
- Co-developed a voice-controlled UR5e arm: a local LLM plans actions; a validator blocks unsafe ones.
- Built perception with no task-specific training data: a VLM finds spoken objects and guides segmentation.
- Served Whisper, Qwen3 and Qwen2.5-VL locally with vLLM; wrote the Python runtime and hardware-mocked tests.
- Validated pick-and-place, swap and stacking on the real arm.
“Most Innovative Project” Jury Award at METU CENG Demo Day 2026




