• Develop mobilenetv3 semantic segmentation model to identify room features (doors, windows, etc) from video, export model with tflite, implement in android on GPU. 50% mIOU, 12ms runtime on Samsung S10.(tensorflow, pixel-level class imbalance, offline augmentation, curriculum learning, quantization, Java)• Develop mobilenetv2 binary semantic segmentation model to identify yoga mats from fixed camera position, export model with tflite, implement in C++ with tensorflow API. 80% IOU on limited test set.(tensorflow, tflite, strong augmentation, pruning, quantization, data and label collection)• Develop 3D skeletal-pose template matching algorithm in C++ for real-time, scalable exercise detection from video stream. 80% Recall. (C++, pose tracking, vectorized math operations, normalization)

• Develop machine learning object detector for camera-based autonomous vehicle feature (tensorflow, keras, cnn, bounding boxes, image data augmentation, transfer learning)• Analyze radar data for preprocessing of machine learning classifier algorithm• Develop state machine to control autonomous vehicle feature• Lead autonomous vehicle feature FMEA through concept and production milestones• Develop electric motor model to investigate motor control algorithms, verify with real-world data, design and implement fuel economy optimizations for hybrid applications• Develop diagnostic to detect PCV system leaks. Build, test, and calibrate to production-ready• Develop vehicle model for concept EV powertrain, implement robustness improvements, verify with SIL• Develop model-based simulation for misfire detection algorithm, verify with HIL