Research under the supervision of Dr. Arman Roohi.

• Crazyflie Isaac Framework (CIF): Developed and open-sourced a modular NVIDIA Isaac Sim framework for Crazyflie research, integrating configurable controllers, sensors, environments, and autonomous algorithms for DRL, multi-drone, benchmarking, and sim-to-real experimentation (GitHub).
• Vision-Language-Action (VLA) for Embodied Robotics (ongoing): Developing an application-specific VLA framework for resource-constrained robots, exploring DINO/SigLIP vision encoders, LoRA-adapted LLaMA models, knowledge distillation, imitation learning, and simulation-based RL for task-specific policy learning.
• Vision-Based Tiny-Drone Navigation: Developed and deployed a fully onboard Crazyflie navigation stack integrating INT8 monocular depth, multi-range sensing, occupancy mapping, and dynamic path planning on GAP8.
• Edge Perception / Temporal Consistency: Developed lightweight TinyML perception for resource-constrained edge platforms, including temporally consistent obstacle detection, embedded inference, BLE, and energy-aware sensing.

Research under the supervision of Dr. Yasemin Ozkan Aydin.

• Learning-Based Quadruped Locomotion: Developed DQN-based locomotion for a salamander-inspired quadruped, integrating spinal–limb coordination and validating policies in simulation and on a physical Dynamixel/Raspberry Pi platform (Demo).
• Autonomous Tunnel Burrowing System: Developed and field-tested an autonomous excavation robot integrating depth perception, force/torque sensing, motion planning, and interchangeable tools (Demo).