Selected results

15.9%lower forecasting errorAverage MAE reduction vs single-source baselines 90%+less formulation timeFrom weeks to days 25%+lower formulation costsAI-optimised product formulations 150+health claims supportedStructured evidence substantiation

Selected work

Research · First-author publication

Forecasting energy use across buildings

15.9% lower mean absolute error

New buildings often lack enough data to train reliable forecasting models. I investigated how transformer models can learn from other buildings.

Combining multiple source datasets reduced error by an average of 15.9% for 24-hour forecasts, compared with single-source baselines using the vanilla Transformer.

Read publication (opens in a new tab)
Methods and citation

Compared PatchTST, vanilla Transformer and Informer across 16 building datasets from the Building Data Genome Project 2. Evaluated zero-shot and fine-tuning approaches; the 15.9% result uses a multi-source zero-shot setup.

Published in Energy and Buildings (Q1 journal, 2025).

First page of Spencer et al. (2025) in Energy and Buildings: title, authors from Massey University, and abstract on transfer learning with transformers for building energy forecasting.
Spencer et al. (2025). Transfer learning on transformers for building energy consumption forecasting—A comparative study. Energy and Buildings, 336, 115632.

Industry · Radix Nutrition

AI for nutrition formulation

From weeks to days

I led the design and deployment of AI systems for nutrition optimisation, combining product ownership with hands-on machine learning and software engineering.

  • Reduced formulation iteration time by over 90%.
  • Co-developed AI-optimised formulations that reduced formulation costs by over 25%.
  • Led evidence substantiation for over 150 general-level health claims for a single product line.
Approach and media coverage

Integrated major food composition databases and regulatory requirements, using the FSANZ framework for claims substantiation. Modelled formulations against nutrition rating and nutrient profiling systems.

Built dashboards, user interfaces and automation adopted across R&D and Operations.

Experience

AI Engineer

Beyond Data · Contract

Jun 2026–Present

Contract role alongside the PhD, working across the full delivery lifecycle — from requirements and solution architecture through data engineering, ML development, and deployment.

Research Assistant · Health Data Science

The University of Queensland · Contract

Jun 2026–Present

Applying clinical language models to large-scale veterinary electronic patient records as part of a natural language processing research project within the School of Veterinary Science.

Tools: Python, Hugging Face Transformers and spaCy.

PhD Researcher · AI for Epidemiology

The University of Queensland · SAAFE CRC

Apr 2026–Present

Developing privacy-preserving digital twins for antimicrobial resistance risk assessment across human, animal and environmental health with SAAFE CRC. Research combines scalable machine learning, synthetic data generation and simulation using UQ’s Bunya high-performance computing infrastructure.

Data Scientist → Senior Data Scientist

Radix Nutrition

Apr 2024–Oct 2025

Joined as the first technical hire and established the data pipelines, development environment and initial machine learning systems. Promoted from Data Scientist to Senior Data Scientist in January 2025, leading AI delivery across R&D and Operations.

See the nutrition optimisation work →

Robert Spencer at Radix Nutrition
Radix Nutrition, 2025

Research Assistant · AI for Energy Forecasting

Massey University · Contract

Sep 2023–Oct 2024

Researched transfer learning for building energy forecasting from September 2023 to October 2024. Compared transformer architectures and evaluated zero-shot and fine-tuning approaches for transferring models between buildings, resulting in a first-author publication in Energy and Buildings (2025).

Robert Spencer with the Massey research team
Massey research team, 2025

Education

My background in health and human performance led me into analytics, software engineering and AI research.

Doctor of Philosophy · AI for Epidemiology

The University of Queensland

2026–2029 · Expected

Research: SAAFE CRC Digital Twins for Antimicrobial Resistance Risk Assessment Modelling.

Fully funded Research Training Program (RTP) Scholarship and SAAFE CRC Top-Up Scholarship. Expected completion: September 2029.

Master of Computer Science · Artificial Intelligence

Monash University

2024–2027 · Expected

Specialising in artificial intelligence, with a High Distinction average (GPA: 3.83/4.0). Completed 9 of 12 units before commencing my PhD; completing the remaining coursework alongside doctoral research. Expected completion: December 2027.

Master of Analytics · Health

Massey University

2023–2024

Awarded with Distinction (GPA: 8.17/9.0), specialising in Health. Awarded a Massey University Research Fund (MURF) Seed Grant to develop machine learning approaches for occupant comfort and energy performance in commercial buildings. Research led to a first-author publication in Energy and Buildings (Q1 journal, 2025).

Robert Spencer at Massey graduation
Master of Analytics graduation

Bachelor of Health, Sport & Human Performance

The University of Waikato

2020–2022

Double majors in Human Performance Science and Community Health — a grounding in health data, stakeholder context, and cross-disciplinary thinking that still informs my technical work.

Robert Spencer at Waikato graduation
University of Waikato graduation

Let’s connect

Based in Brisbane. I welcome conversations about practical AI systems, research and collaboration across industry, health and the environment.

Email meLinkedInDownload CV