- Statistical Learning and Data Mining: Instruct students on implementing feature engineering via pandas, plotly, numpy, matplotlib, sklearn and implement algorithms such as linear regression, logistic regression, generalised additive model, xgboost, neural network.- Predictive Analytics: Instruct students on implementing time series forecasting such as time series decomposition, Holtwinter, SARIMA and recurrent network.

Education Sector: Developed a skills matching platform for the Vocational Education sector, using Weaviate and ElasticSearch, integrating Generative AI models and tools like GPT-4, and Hugging Face and Langchain to automate training course development. This innovation has streamlined workflows and redefined approaches to education and Recognition of Prior Learning (RPL), enabling efficient skill mapping for RPL.Music Industry: - Automated the transformation of 30 million musical works into a denormalized format using Apache Spark, enhancing a high-performance data pipeline for music record matching in an AWS environment. - Led R&D efforts to evolve beyond traditional data processing methods, applying techniques like locality-sensitive hashing and NLP vector embedding, resulting in a scalable, accurate matching system that significantly improved processing times, reduced costs, and increased match rates.Logistics Industry: Utilising gradient boosting algorithms and Shapley explanatory models applied a methodology fine tuned in previous projects to explore and discover key business drivers for a logistics technology company. The various ML models were developed into a pricing recommendation application. This application estimates price sensitivities and recommends prices on both the buy and sell side of a logistics market place based on key attributes of each logistics job.

Healthcare Sector: - Developed a novel approach to rapidly analyse a massive dataset describing every emergency department and hospital admission over a 5 year period from a population of 2 million patients in Melbourne's largest and most diverse health district. Utilising gradient boost machines (XGBoost) and Shapley explanatory models, the methodology was able to identify key drivers to various hypotheses out of 100s of data fields. Specifically, this resulted in understanding key drivers of diverse problems including drivers of cardiovascular disease and length of hospital stay for unexpected hospital admissions.- Developed a proof of concept real-time tool to predict length of stay in Emergency Departments. Utilised an LSTM neural network to predict demand within the ED based on data readily available from ED admin systems.- Led the development of a real-time student wellbeing monitoring system.ML Application Deployment:Led the end-to-end deployment of dashboards and applications on AWS, including the set-up and configuration from scratch using AWS EC2, Elastic IP, along with Dash, Docker-Compose, Nginx, and Certbot for rapid prototyping and deplyment.

Focus on deliver solutions for R&D projects such as:• Collect data from IoT trackers to analyze network coverage and user's behaviour via statistical hypothesis testing and unsupervised clustering.• Leverage a clustering technique to provide a smart data compression for WiFi geolocation service using Sigfox communication. In a simulation environment, the algorithm achieves the same performance as a traditional method in 99% of cases while only using 1/4 of the packet size.• Build a computing model based on Monte Carlo to analyze the Sigfox network capacity. Since directly applying the Poisson distribution to compute the packet loss rate is complicated in the context of the Sigfox technology, the proposed model avoids that by directly simulating the transmission scheme.• Apply a distributed optimization technique to address a package scheduling problem for a streetlight control application via Sigfox technology, in which an integer optimization problem of overlapping transmission is addressed.• Play a role of a key firmware developer as well as QA process designer for a mainstream product (Xkit) which was successfully launched in early 2017.Besides, I am also involved in:• Writing firmwares and Q/A process at factories.• Provide support and consultancy to clients to develop their own PoC.

Research Profile: https://www.researchgate.net/profile/Ho_Tam• Developed an optimization framework based on Successive Convex Quadratic Programming to extensively solve challenging problems in telecommunication world such as: 1. Nonconvex objective functions involving logarithmic functions 2. Mix-binary problems 3. Nonconvex constraints• Performed numerical simulations for model validation in Matlab and published the findings in 8 journals and 10 conferences (4 journals published in the top 10 journals in the telecommunications area).

- Implement a TCP server to gather data from sensor nodes (Raspberry Pi).- Implement occupancy detection based on RSSI of Bluetooth signals for lighting application on Python.

- Design algorithm to extract fetal EEG signal from a maternal using subspace projection techniques.- Design a routing protocol on application layer of Zigbee for Automatically meters reading application.- Implement protocol on CC2430DK evaluation kit of Texas Instrument.

- Apply linear phase space projection technique to separate the fetus EEG signal from the maternal EEG signal which are chaotic signals.