曾微中 Email & Phone Number

CAREER SUMMARY:Experienced in modeling functional behavior, system-level traffic, detecting bottlenecks, and providing solutions.Integrating complex operations into a well-organized system.Familiar with CNN models and inference frameworks.---Experienced in :- Standalone performance simulator for architecture exploration.- Modeling multi-core system of AI accelerator. (TLM / Systemc)- Analysis performance & power for AI accelerator.- Functional model for CIM(compute in memory).- Design verification & verification tools- Build benchmark to test for unexpected detail of hardware architecture from our competitor.Machine learning：Speedup CNNs inference for edge device -Design a C-based Inference framework. (基於C語言 的輕量AI框架 ) --For resource limited device --Code size 158k --Compilation time 4sec --C only -Fixed-point quantization without retraining (定點數量化CNNs網路) --Power reduction over 92% --mode size reduction over 55%~75% --little accuracy dropWafer Inspection with DNN model.(晶圓錯誤辨識模型)Build a fast object tracking model with binary operation. (Binary weight 快速物件辨識模型)Distributed DNN research.(分散式AI模型)相關工具:Linux environment, Darknet, TensorFlow, Caffe, memory leak tool, code coverage tool, GDB, perf, git, 使用語言:C, C++, JAVA, bash, python, javascript, html, R, 開發物聯網應用整合OM2M , DICOM , 藍芽室內定位 , smart glass