With over 30 years of embedded design experience in DSP and FPGA, my career began in ASIC design and has evolved across SoC platforms; today my focus is AI pipelines, Embedded Vision, and Robotics at the edge.
For the past several years I have benchmarked edge AI accelerators across silicon families, publishing methodology and results.
I have explored the use of computer vision and agentic AI approaches to help humans interact with robotics.
When not working, I am a passionate rock climber and woodworker.
Methodology-first, vendor-neutral evaluation of edge AI accelerators — power, throughput, latency, energy-per-inference, accuracy retention, and pipeline-level performance on cascaded real-world workloads. Reproducible measurement, published methodology, head-to-head comparisons.
Edge AI deployment from custom datasets to hardware-accelerated inference. End-to-end work spanning model exploration, quantization-aware training, and deployment on external AI accelerators (AzurEngine, Axelera, Hailo, MemryX, DeepX), and internal NPUs (AMD Vitis-AI, Qualcomm Hexagon NPU). Hands-on experience with classification, object detection, image segmentation, 3D point-cloud detection, and stereo neural inference. Expertise in building advanced pipelines such as person tracking, pose estimation, face recognition, and hand gestures.
In depth experience in building image-capture pipelines for AMD programmable logic platforms (Spartan-6, Zynq-7000 SoC, Zynq-UltraScale+, Versal AI Edge), including camera calibration and ISP tuning, for mono, dual(stereo), and multi-camera systems.
Combining computer vision and AI to make robots respond to humans and their surroundings. Implemented hand-controlled robotic arms and mobile robots using MediaPipe, pose estimation, and ASL recognition. Researched LLM-based agentic AI in ROS2 for autonomous robot control. Created digital twins environments for closed-loop simulation and validation.
I am also known as AlbertaBeef