I build Robotics AI systems that combine learning-based policies, Computer Vision, and Deep Learning with efficient Edge AI deployment on Embedded Systems.
My focus is imitation learning (diffusion policies, implicit behavioral cloning), Vision-Language-Action (VLA), sim-to-real transfer, model optimization, and turning research into reliable real-world robotic products.
Reach out at vydat1702@gmail.com or connect via social profiles.
Education
Bachelor's degree, Embedded Systems and Internet of Things
2022 - 2026
Experience
Bosch Global Software Technologies Vietnam
Engineered Edge AI solutions for resource-constrained embedded systems, optimizing deep learning models to ensure efficient real-world deployment. Accelerated on-device inference performance by adapting complex network architectures for hardware-specific constraints.
Ho Chi Minh City
Built an automated benchmarking pipeline across embedded devices, identifying critical bottlenecks to drive data-driven hardware and software optimizations. Executed real-world integration testing and authored technical reports, reducing deployment regressions and guiding product version upgrades.
Deployed TensorFlow Lite models on STM32 microcontrollers using quantization and pruning techniques, significantly reducing inference latency and energy consumption. Engineered an end-to-end IoT data pipeline (I2C/SPI to MQTT) and a real-time monitoring dashboard, ensuring reliable sensor acquisition and accelerating system validation.
Publications
01
Dat Nguyen-Lam Vy, Nghe-Nhan Truong, Van Dang Nguyen, Dao Tuan Dat Vo, and My-Ha Le
Proc. 8th GTSD, Ho Chi Minh City, Vietnam, 2026
02
Dao Tuan Dat Vo, Nghe-Nhan Truong, Dat Nguyen-Lam Vy, Van Dang Nguyen, and My-Ha Le
Proc. 8th GTSD, Ho Chi Minh City, Vietnam, 2026
03
Van Dang Nguyen, Nghe-Nhan Truong, Dao Tuan Dat Vo, Dat Nguyen-Lam Vy, and My-Ha Le
Proc. 8th GTSD, Ho Chi Minh City, Vietnam, 2026
04
Dao Tuan Dat Vo, Van Dang Nguyen, Lam Vy Dat Nguyen, Hao Minh Nguyen Tran, Vo Thi Xuan Nhung, Tran Kim Toai, Nguyen Quoc Khanh, Tran Huu Nhat Thao, and Minh Huan Vo
Proc. 8th GTSD, Ho Chi Minh City, Vietnam, 2026
Projects
Built a Neural Kalman Filter with temporal modeling to estimate vehicle lateral velocity and deployed it on embedded hardware for robust performance on challenging driving conditions.
Built an end-to-end edge AI module transforming sensor signals into spectrum images for machine health diagnosis with optimized model inference.
Re-implemented Diffusion Policy for the Push-T benchmark with a Conditional U-Net 1D, cosine noise schedule, EMA, and DDIM sampling, preserving multimodal action distributions instead of collapsing to a mean.
Re-implemented Implicit Behavioral Cloning as an energy-based imitation learning method and compared it against an MSE baseline on multimodal Particle 2D and block push tasks.
Certificates