Dr Yongchao Huang

Biography

2013-2017, DPhil Engineering Science (Oxford)

2016-2017, MLRG (Oxford), Bayesian/Markovian

2017-2019, Senior Data Scientist in Actuarial industry (UK), machine learning/data science

2019-2022, Postdoc in CS (Oxford), machine learning/reinforcement learning/Gaussian process/MCMC

2022-, Visiting lecturer in CS (Westminster), student supervision

2022-2023, Senior Research Associate in MLG/CBL (Cambridge), probabilistic machine learning

2023-2024, collaborator, MLG/CBL (Cambridge)

2024-2026, early career academics panel, CSIC Engineering department (Cambridge)

2023-, Lecturer/Assistant Professor in CS (Aberdeen), machine learning

I did DPhil in solid mechanics from Oxford during 2013-2017, investigating signal processing (FFT) of impact wave propagation in solids, under supervision of Prof. Clive Siviour. I also spent time in final year 2016-2017 in the Machine Learning Research Group at Oxford, working with Prof. Stephen Roberts on Bayesian data analysis for financial markets. Later I went to the UK actuarial industry, implementing ML algorithms in a commercial environment on corporate scale, first as a senior pricing analyst, then sr. data scientist/manager. I returned to academia after 2.5 years in industry, with keen interest in fundamental research (after attending years of full-suite, classroom ML courses in 2015/2017, 2018/2019 at Oxford), and did 3-year postdoc in Oxford Computer Science, working with Prof. Alessandro Abate on ML, RL and GP for smart energies, during which I also spent some time in the Torr Vision Group as a RA, and as a research software engineer in the Computational Biology Group, working on RL and MCMC methods. Later I did a second postdoc (SRA) in the Machine Learning Group at Cambridge CBL, working with Prof. Zoubin Ghahramani and Dr. Hong Ge on probabilistic machine learning, where I stayed as a collaborator till late 2024. I serve on the early career academics panel at CSIC Cambridge from Sept. 2024. I am a lecturer at Aberdeen from Aug. 2023.

External Memberships

Service to community:

Conferences & Journal review / PC / Organisation: NeurIPS (2026), ICML (2026), AIBSD (2026), Bioinference (2024), ECAI (2024), Theoretical Biology (2024), Neural Networks and Learning Systems (2021), etc

Others: Chartered Management Institute (Level 5, 2019), Royal Statistical Society (RSS, 2023), International Linear Algebra Society (ILAS, 2024/25), Sr. Scientific Advisor to a UK firm (2025/26), official nominator VinFuture prize (2026), FHEA (2026-), etc.

The above list of professional memberships is exhaustively enumerated to the best of my knowledge in this universe as of 2025, any role claimed beyond these is invalid.

Non-professional membership include fans of classics and modern music, e.g. JS Bach (bwv 639), Chopin (Nocturne C sharp minor), heart full of soul (Yardbirds 1965), stairway to heaven (MSG 1973), Lithium (Reading 1992), art of life (Tokyo Dome 1993), OK computer (paranoid android 1997), TMGE (Fuji Rock '98), Jay Chou (Fantasy 2001), whatsername (Argentina 2011), good boy (MAMA 2014), love shot (2018), BTS (airplane pt.2), BP (forever young 2018), GD (drama 2025), and many others. I used to publish original songs and novels some decade ago - both the author and creates are technically boring however fun by nature. 

Research Overview

Broadly classic and modern approaches, theories and applications. Typically,

- Physics-motivated machine learning, science for AI. 2019 - present

- Bayesian, MCMC, variational inference. 2016 - present

- Generative modelling (score-based). 2022 - present

- NeuroML, neural & biological systems. 2024 - present

- Small language models and uncertainties, 2025 - present

- DL, RL, dynamical systems (differential equations - applied maths). 2013 - present

- Interdisciplinary & AI for science. e.g. vision, biology, mechanics, engineering materials/structures, energy, environment, climate, finance, etc.

A large portion of my work centers around log p(x) and its gradients, i.e. how to effectively infer a density and/or efficiently sample it. These trigger some relevant topics related to maths, statistics, computation, and cross boarders. Empirical work involves theoretical and practical elements. 

During my career, I am fortunate to have worked with many great minds in several fields including applied maths, computer science, physics, mechanics, finance, civil and information engineering, on both theories and applications. This cross-boundary journey is very much inspired by Shannon, Feynman, Heisenberg, Grigori Perelman, James Simons and David Mackay, among many others. If I am to pick 1 or 2 favourite subjects, it would be numerical linear algebra or computational stats - for many years I have been thinking about matrices and their representations, and a long journey to explore still.

Having worked in both academia and industry, and learning from great minds, I am fortunate to be able to read, write, derive, code and assess independently, through years of practice. For over a decade, I have been doing reading, writing and coding on a daily basis. Frankly speaking, it would be a shortcut, and less challenging, for me to apply SOTA methods and experiences to downstream tasks, as I come from an interdisciplinary background. But I chose to go hard, deep and fundamental. In my own naive view and interest, creating is preferred than following or applying.

Following the principle 'information is physical' (Landauer, 1991), I run the 'Computational and Physical Learning' lab (CPL, 2023-present) at Aberdeen, being the founder and solo member, aiming to make fundamental contribution to the arena of machine intelligence, e.g. improving existing methods or inventing new - making tools and building bricks purely driven by curiosity. With very limited resource and energy to build a team (and a website), I therefore mostly do all the work end-to-end and write papers on my own. In light of the absence of a dedicated statistics unit, I also founded the Probabilistic Machine Learning (PML) Group in 2023 to establish the university’s first research team focused on the probabilistic, statistical, and mathematical foundations of AI.

Current Research

Bayesian maths, sampling methods, applications.

Past Research

Differential equations, signal processing, surrogate models, learning-based and physics-based modelling.

Knowledge Exchange

Welcome academia/industry collaborations, knowledge exchange and public engagement.

Teaching Responsibilities

Fellow of higher education academy (FHEA, awarded 2026). Having taught following courses in the past:

1. Lectures: <Computational Intelligence>, Aberdeen, co-coordinator/designer 2024 & co-coordinator 2025

2. Lectures: <Machine learning>, Aberdeen, co-coordinator 2026

3. Lectures: <Python programming foundations>, Aberdeen, co-coordinator 2025

4. Practicals: <Object-oriented programming>, Aberdeen, co-coordinator 2026

5. Lectures: <Distributed systems>, Aberdeen, co-coordinator 2025

6. Lectures: <Introduction to Software Engineering>, Aberdeen, co-coordinator 2023 & coordinator/designer 2024

7. Lectures: <Software Process and Management>, Aberdeen, co-coordinator 2024 & co-coordinator 2026

8. Practicals: <3f8: Inference>, Cambridge, demonstrator 2023

Others: Linear Algebra, ML, Engineering Maths (summer schools).