Email: kvali [at] ucdavis [dot] edu
I am Kourosh Vali, a Ph.D. student in the Electrical and Computer Engineering Department under the supervision of Prof. Soheil Ghiasi. I received my Bachelor of Science in Electrical Engineering from the University of Tabriz in 2018. I joined the Laboratory for Embedded and Programmable Systems at UC Davis the following year, where I am a Graduate Student Researcher ever since.
My research is on designing Embedded software system solutions for health applications. I’m currently leading the Transabdominal Fetal pulse-Oximetry system (TFO) project, where I mainly focus on system design, signal processing, and machine learning to estimate the blood oxygen saturation.
EEC 284 Optimization of Embedded Computing Systems FQ19
EEC 007 Introduction to C and microcontrollers – WQ19
EEC 018 Digital systems design – SQ19, SQ20
EEC 172 Embedded Systems – WQ20
A cool project by my smart students:
Software and System design internship at Nemone Pardaz Azar Co. Summer 2017
- Machine Learning
- Embedded Systems
- Data Analysis in Health Applications
- Data Science
- Signal Processing
- Hardware and Software Co-Design
- System Design and Optimization
- Internet of Things
- Computer Vision
- Parallel Computing
- Kourosh Vali, Begum Kasap, Weitai Qian, Christina M. Theodorou, Tailai Lihe, Daniel D. Fong, Christopher D. Pivetti, Edwin Kulubya, Kaeli Yamashiro, Aijun Wang, M. Austin Johnson, Herman L. Hedriana, Diana L. Farmer, Soheil Ghiasi, “Non-invasive Transabdominal Assessment of In-Utero Fetal Oxygen Saturation in a Hypoxic Lamb Model”, Abstract Accepted for Annual Pregnancy Meeting of the Society for Maternal-Fetal Medicine, 2021
- Daniel D. Fong, Kaeli J. Yamashiro, Kourosh Vali, Laura A. Galganski, Jameson Thies, Rasta Moeinzadeh, Christopher Pivetti, André Knoesen, Vivek J. Srinivasan, Herman L. Hedriana, Diana L. Farmer, M. Austin Johnson, and Soheil Ghiasi. “Design and In Vivo Evaluation of a Non-invasive Transabdominal Fetal Pulse Oximeter“, IEEE Transactions on Biomedical Engineering, 2020
- Daniel D. Fong, Kourosh Vali, Soheil Ghiasi, “Contextually-aware Fetal Sensing in Transabdominal Fetal Pulse Oximetry,” ACM/IEEE 11th International Conference on Cyber-Physical Systems (ICCPS), 2020
- Daniel D. Fong, Kaeli J. Yamashiro, M. Austin Johnson, Kourosh Vali, Laura A. Galganski, Christopher D. Pivetti, Diana L. Farmer, Herman L. Hedriana, Soheil Ghiasi. “Validation of a novel transabdominal fetal oximeter in a hypoxic fetal lamb model“, Reproductive Sciences, 2020
- Daniel D. Fong, Kaeli Yamashiro, M. Austin Johnson, Kourosh Vali, Laura Galganski, Christopher Pivetti, Diana Farmer, Herman Hedriana, Soheil Ghiasi, “Validation of a Novel Transcutaneous Fetal Oximeter in a Hypoxic Fetal Sheep Model”, Abstract for Annual Pregnancy Meeting of the Society for Maternal-Fetal Medicine, 2020
- Daniel D. Fong, Vivek J. Srinivasan, Kourosh Vali, and Soheil Ghiasi. “Optode Design Space Exploration for Clinically-robust Non-invasive Fetal Oximetry“,ACM Transactions on Embedded Computing Systems (TECS), Volume 18, 2019
- Daniel D. Fong, Vivek J. Srinivasan, Kourosh Vali, and Soheil Ghiasi. “Optode Design Space Exploration for Clinically-robust Non-invasive Fetal Oximetry,” presented at Embedded Systems Week – Hardware/Software Codesign and System Synthesis (CODES+ ISSS), 2019
- COVID-19 cases in the US and social-distancing measures (Supervised by Prof. James Sharpnack – STA 208)
• We used the social-distancing data provided by SafeGraph, and Designed ANN models to predict the actual incident rate (# cases per 100,000 people) in the last two weeks with the MAE of 71 cases on the COVID-19 US daily count data for each state. We could also anticipate 15 out of the top 20 states with the highest incident rate changes after this period.
- Predicting Length of Stay for Patients in ICU (Supervised by Prof. Soheil Ghiasi – EEC 289Q)
• We explored the Machine learning models to predict the length of stay of a patient using the initial information and diagnosis to help resource allocation to ICU patients using the MIMIC-III dataset. The best result was achieved with an ANN with one hidden layer and 100 parameters. The model could predict the length of stay with the MAE of 137.82 hours.
- Neural Decoding of a Rodent on a rotating platform (Supervised by Prof. Karen Moxon – MAE 289)
• An implementation of the Kalman filter estimation for neural decoding of a rat’s limbs tilted on a platform by having the Ground Reaction Forces of the platform. We found out that the best estimator decoded the activity of Limb 3 for fast clockwise rotations of the rat with 5ms bin sizes with the MSE of 10.7.
- A CUDA-based Batcher-Banyan Network for Bitonic Sort (Supervised by Prof. John Owens – EEC 289Q)
• We described the Batcher-Banyan sorting network, a type of bitonic mergesort on the CUDA library. This sorting algorithm uses Butterfly and Shuffle blocks for this task. By implementing these blocks in the GPU shared memory space, we could improve upon the throughput of Radix sort by an average of over 60 % in sorting less than 213 random elements.
- Enhancing Sentimental Analysis of User Reviews on BERT (Supervised by Prof. Ilias Tagkopoulos – ECS 289G)
• We Improved the Sentimental Analysis of the movie reviews of the IMDb Dataset on the Google BERT NLP model using Transfer Learning. We employed a Large number of location reviews from the YELP Dataset to refine the reviews’ predictions. We managed to increase the AUC by 0.5% to 89.3%.
- TD Learning to Find Car Control Policy (Supervised by Prof. Robert Cui – EEC 289A)
• We employed TD learning to train a graphically-simulated car to navigate through a closed track. The prior information is updated in each step from the distances from the walls and the car’s speed. The best performance was observed with 64 step TD learning and after 200 episodes of training and manages to circulate an average of 3 laps on the track with no collision.
- Improving a Panoptic Image Segmentation Model (Supervised by Prof. Zhi Ding – EEC 206)
• We investigated an algorithm to merge the DeepLab V2 Semantic Image Segmentation model and the MaskRCNN ResNet Instance Image Segmentation model to define the Panoptic Image Segmentation model and achieved a Panoptic Quality score of 38.1. This project won the competition in a poster presentation at UC Davis.
- Speech English Accent Transformation (Supervised by Prof. Zhi Ding – EEC 201)
• We developed an algorithm to classify the speech sounds of spoken British, Midwestern, and Western Accents with 88.14% accuracy and Transform them into the accent of interest by mapping the Mel-Frequency Cepstrum Coefficients and tuning with Dynamic Time Warping. The final speech is Synthesized using LPC vocoders.
- Real-time Heart Rate Sensing of two persons (Supervised by Prof. Soheil. Ghiasi – EEC 284)
• We developed an embedded software system to detect and differentiate two person’s heartbeat with a single sensor on an ARM microcontroller. The reliability of the system was improved by reducing flicker noise from the system through Synchronous Detection.
- Ball tracking using a monocular camera on a robotic arm (supervised by Prof. Ghader Karimian)
• We implemented an algorithm to track a ball and catch it with a single USB webcam attached to a robotic arm based on Kalman Filters on Raspberry Pi. This improves ball detection and tracking speed 3x over hough transform.
- Path follower robot with a webcam on Raspberry PI (supervised by Prof. Ghader Karimian)
• We Designed a robot to trail a path with a USB webcam on Raspberry Pi in real-time.
- The UART module implemented on an Altera FGPA (supervised by Prof. Ghader Karimian)
• We implemented the UART protocol on an FPGA that allows for various baud rates, data bits, and parity control.
2236 Kemper Hall
One Shields Avenue, University of California Davis