Programme Overview
I'm doing an MSc in Biomedical Engineering at Imperial, on the Computational Bioengineering track. It suits me because it moves cleanly between engineering, biology, and computation instead of treating them as separate worlds.
Autumn Term
Systems Physiology
- Mathematical modelling of organ systems including the kidney, heart, lungs, muscles, digestive system, endocrine system, and nervous system.
- Taught from an engineering and biophysics angle rather than descriptive anatomy.
Data Analysis for Research
- Statistical methods for bioengineering research, implemented in Python.
- Included hypothesis testing, ANOVA, Bayesian inference, generalised linear models, bootstrapping, Monte Carlo methods, and PCA.
- The useful part was not just running methods, but understanding when they should and should not be used.
Digital Biosignal Processing
- Discrete-time signal processing for biomedical data.
- Covered sampling theory, DFT, z-transform, FIR/IIR filter design, and power spectral density estimation.
Reinforcement Learning for Bioengineers
- Covered MDPs, the Bellman equation, dynamic programming, tabular RL, and deep RL.
- Worked through continuous and high-dimensional action spaces in Python and PyTorch.
- Applied in robotics and neuroscience settings.
- Final project: ProteinTuneRL: A Reinforcement Learning Perspective on Antibody Design.
Spring Term
Brain Machine Interfaces
- Covered ECoG, EEG, MEG, their trade-offs, and decoder theory.
- Looked at clinical uses such as restoring movement in paralysed patients and deep brain stimulation for Parkinson's.
- Included MATLAB lab work on decoding neurophysiological signals.
Artificial Intelligence for Drug Discovery
- Covered the drug discovery pipeline, chemoinformatics, and bioinformatics basics.
- Spent a lot of time on where AI goes wrong: inflated benchmarks, class imbalance, sparse data, and misleading evaluation.
- Useful mainly because it forced a more critical reading of biomedical AI claims.
Image Processing
- Image transforms, neighbourhood operators, segmentation, registration, and image synthesis.
- Applied across clinical imaging and research problems.
Biomimetics
- How biological solutions inspire engineering, from the molecular scale up to full systems.
- Covered bio-inspired adhesives, materials, sensing systems, structural colour, and parallels between biological and artificial vision.
Medical Device Certification
- EU regulatory framework, CE marking, and device classification.
- Safety and hazard analysis, risk management, product development lifecycle, and intellectual property.
MSc Journal Club
- Weekly critical review and discussion of current papers.
- Mostly useful for presenting, defending, and arguing through primary literature properly.
Individual Project
Physics-Informed Operator Learning for Microbial Dynamics
Supervisor: Prof. Reiko Tanaka · Thesis submission: September 2026
The dissertation is about whether a general model can infer hidden microbial-system parameters directly from observed trajectories, instead of retraining a separate physics-informed model for every new dataset. The problem sits in the generalised Lotka-Volterra setting and is motivated by the fact that microbial ecosystems are nonlinear, messy, and hard to model well with limited real data.
I'm comparing three approaches: a baseline PINN retrained per dataset, PI-DION as the main inverse-operator approach, and an exploratory QPINN setup in PennyLane. The broader question is whether operator learning can generalise better when data is scarce and the system varies across datasets.
Generalisation is being tested on real mouse gut microbiome data as well as biologically constrained synthetic data. Early results suggest that the inverse setup works well on synthetic systems, but the main challenge is still the jump from synthetic to real data, which is really the whole point of the operator-learning approach.