Po-Yu Chen Email and Phone Number

[Lead the Development of Core Models and Research]1) Decision Making: Developed the next-best action engine using Neural and Gaussian Process-based Multi-Armed Bandit.2) NLP models: SOTA research in LLM fine-tuning and RL-guided LLM.3) Generative models: SOTA research in differentially private image and multivariate time-series generation. 4) Reinforcement Learning: Developed Multi-agent RL models that solve macro/microeconomic problems. 5) Representation Learning: Developed new user embedding that summarises customers' behaviours with SOTA ML techniques, including contrastive learning and self-supervised approaches.5) Sustainable AI: Develop a new framework that takes Model Utility, Privacy, Fairness, Greenhouse Gas Emissions, and Model Explainability into account.6) Established new model deployment standards and internal benchmarks.[Key Use Cases]Customer Behaviour Analysis (e.g., Churn and Silent Leavers), Propensity and Forecasting Models, NLP Models, Causality Analysis, Generative Modelling, Reinforcement Learning. [Other Contributions]RnD lead, Tech training lead

[Assistant Supervisor]supervised PhD students whose research is related to anomaly detection in distributed networked systems and ML-aid RF sensing.[Machine Learning]investigated differential privacy and membership inference attacks in machine learning models. I also investigated adaptive ML models and self-supervised embeddings for device-free Radio-Frequency (RF) sensing, including human activity recognition (HAR).[Networking Protocols]proposed anomaly detection and task scheduling protocols for distributed multi-hop networks and low-power wide-area network (LPWAN) systems.[IoT/CPS System Development]experienced in developing and deploying wireless sensor networks (WSNs), collecting real-time streaming data in the UK vineyards and wineries.Teaching - co-designed the tutorials and demos for the Pervasive Computing course in the Computing Department of Imperial College.[Public Engagement]presented research works in Google and Imperial College poster competitions, Imperial Festival and Science Museum Lates to the public.[Research Projects]experienced in multiple national and national projects, including EPSRC WEFWEB (2016-2018), NEC (2018-2019), PETRAS SmartPort 4.0 (2022-present) and Alan Turing Institute sustainable infrastructure (2019-2022).

Big data analytic (e.g. anomaly detection, pattern recognition, data property extraction) in distributed systems.

[Sustainable Infrastructures]improve data reliability with new ML and data analytic tools for failure detection, diagnosis, and recovery.[Machine Learning]feature extraction, latent-space embedding and self-supervised learning models for multi-modal time-series data steaming.[Network Anomaly Detection]optimal and adaptive monitor placement for network tomography.

Wireless sensor network system and platform development and consultant.Smart home platform development.