Xian Long

1. Spearheaded the development and maintenance of scalable AI infrastructure, ensuring high availability and performance for mission-critical AI applications. 2. Designed end-to-end MLOps pipelines, automating model training, deployment, and monitoring, resulting in a reduction in deployment time and a continuing increase in model accuracy. 3. Regularly assessed and implemented state-of-the-art AI research to optimize model performance and maintain competitiveness in the rapidly evolving AI landscape.

Active learning Designed and implemented an end-to-end active learning pipeline that selected the most informative data for improving the model performance. This MLops pipeline automates active selection, data augmentation, model training, model evaluation, and model update with minimal human intervention. Other Projects: • Design an impaired camera detection system including blur, haze, and dirt detection functions. • Research and develop an LLM-based AI agent to gain insights into the internal data. • Research the latest zero-shot, few-shot learning algorithms. Presien is a global AI vision company improving how heavy industry goes to work. We turn inputs into intelligence so heavy industry can see the path to progress. https://www.australianmining.com.au/position-partners-covers-all-angles-with-blindsight/

Face recognition module (access control and monitoring systems) • Train high-performance face recognition models with the latest network structures • Build automatic data fusion pipelines to create million-level datasets • Develop a high-speed high-dimensional search engine for face matching (10 times faster) • Implement GAN-based models for image enhancement (HDR and super-resolution) • Develop face anti-spoofing and face tracking modules. Object detection and video analytics • Train the latest object detection/tracking models for different applications such as persons, vehicles, and PPE based on customers’ needs • Customize and deploy the models on the cloud and different edge devices • Develop the models for video motion detection and video anomaly detection NLP and Chatbot system • Test the latest NLP algorithms including BERT and GPT-series for different tasks • Build a chatbot pipeline including audio-to-face motion and video rendering engine Food intake measurement system • Develop multimodal algorithms (unsupervised and supervised) for food segmentation • Estimate the volume by combining the segmentation and the depth (LiDAR) information

As part of my PhD research, I am the architect and leading developer of multiple software to analyze neural activity data, revealing the dynamical principles of neural patterns and improving classification accuracy of different recognition stages. • Capable of analyzing terabytes of data • Object-oriented • Parallelization (batch scripts) • Over 200 post-analysis and visualization functions • Implementation of a wide range of signal processing (Hilbert, Wavelet, EMD), computer vision (optical flow, feature extraction), machine learning (regression, PCA/ICA/Isomap, Kmeans, SVM, Graph theory, HMM, MCMC, statistical analysis/modelling) and deep learning algorithms (CNN, RNN, SNN, LSM, physics-inspired NN) • Software used by more than 15 team members

• Development of novel data analysis method for studying the dynamical patterns in the large-scale human cortex data. • Application of deep neural network and unsupervised methods to automatically detect neural patterns. • Professional writings and advanced graphics visualisation using Adobe Illustrator. • Assistant in supervision of eight projects (Honours, Summer and Talented Student Program (TSP) projects).

rial/laboratory demonstration and course material preparation for ELEC5305 - Acoustics, Speech and Signal Processing. The course covers acoustic signal processing, pattern recognition, machine learning, and deep learning. Teaching experience with CNN and RNN implementation from scratch and with Tensorflow and Keras and other machine learning algorithms from scratch., such as kmeans, HMM, and EM-GMM.

Development of a topological model using large deformation diffeomorphic metric mapping for the head-related transfer function (HRTF, for 3D audio individualization, used in spatial audio and virtual reality), resulting in the prediction of HRTF using 3 principal components (kernel PCA).

• Quality control and supervision of image annotation tasks. • Annotate facial images and video sequences with landmark points for use in deep learning algorithms. • Provide feedback on software usage and annotation guide.

• Annotate facial images and video sequences with landmark points for use in deep learning algorithms. • Collect data for annotation. • Provide feedback on software usage and annotation guide.

customer service

As part of my Master research, I am the architect and developer of advanced signal processing and image processing algorithms to improve the resolution of ultrasound images.

• Advanced data analysis on terabytes of neural data using signal processing, ML, statistics, and interdisciplinary ideas from computer vision, turbulence physics, statistical physics, ecology and finance. • Development of software (biomarkers) to classify different recognition tasks and sleep stages, achieving an ~15% increase in accuracy and ~70% reduction of data used for classification. • Diverse modelling for predicting neural dynamics, such as neural network models, statistical models, and nonlinear dynamical models. • Development of a topological model using LDDMM for the head-related transfer function (HRTF, in virtual reality), resulting in the prediction of HRTF using 3 principal components. • Development of software - second harmonics imaging - to improve the prediction accuracy (~20%) of bone intensity measurement in ultrasound images.

• Development of software using the array signal processing alogrithm - MVDR with phase smoothing - to improve the resolution (~10% increase) of the biomedical images. • Development of software using the image processing algorithm - beamforming with phase compression - to improve the contrast of the biomedical images.

Thesis: Development & Optimization of Microwave Dielectric Materials

Electronic Science and Technology with an photoelectric option