Daniele Pessina
- Time-series modelling in data-scarce regimes with advanced machine learning architectures like neural ODEs, transformers, autoencoders with transfer learning and reinforcement learning methods.
- Probabilistic programming and uncertainty quantification of mechanistic, hybrid and black-box models, enabling robust parameter estimation and decision-making under uncertainty.
- Differentiable programming for process modelling and simulation, delivering GPU-accelerated, autodiff-compatible frameworks in Julia and JAX.
- Hybrid modelling and ML-assisted sensitivity analysis of large, non-linear pharmacokinetic models, improving interpretability and identifiability.
- Development of ML-based soft sensors for real-time product quality monitoring in manufacturing, reducing reliance on offline batch testing.
- Experimental validation of modelling hypotheses through medium-scale protein crystallisation studies using MT EasyMax and PAT tools (FTIR, UV-Vis, image analysis).
-
Pessina D.; Tian, T.; Watson, O.; Heng, J. Y. Y.; Papathanasiou, M. M. (2025) Transfer Learning Assessment of Data-driven Crystallisation Processes via Constrained Neural Ordinary Differential Equations
-
Pessina, D.; Calderon De Anda, J.; Heffernan, C.; Heng, J. Y. Y.; Papathanasiou, M. M. (2025) Integrated In Vitro/In Silico Uncertainty Quantification Method for Protein Crystallization Models
-
Pessina, D.; Calderon De Anda, J.; Heffernan, C.; Tian, T.; Watson, O.; Heng, J. Y. Y.; Papathanasiou, M. M. (2025) Model-based approach to template-induced macromolecule crystallisation
-
Pessina, D.; Abbiati, R. A.; Manca, D.; Papathanasiou, M. M. (2025) Machine learning-enhanced Sensitivity Analysis for Complex Pharmaceutical Systems
- CriSTool - the Crystallisation In-Silico Toolbox in Julia CriSTool is a Julia toolbox for simulating crystallisation population balance models, estimating kinetics from data, and exploring uncertainty and sensitivity.
- Crystalline - Neural ODEs for Transfer Learning of Crystallisation Processes Research code for training Augmented Neural ODE (AugNODE) models on crystallisation kinetics using JAX, Equinox, and Diffrax. The package includes training utilities for fresh and transfer‑learned AugNODEs plus a Dash dashboard to launch runs and inspect plots interactively.
About me:
I am a final year PhD student at the Sargent Centre for Process Systems Engineering in the Department of Chemical Engineering at Imperial. I am supervised by Dr. Maria Papathanasiou and Prof Jerry Heng, and my PhD is funded through a CASE studentship with AstraZeneca.
My research is focused on template-induced biomolecule crystallisation modelling and process design. I work at the interface of process systems engineering, machine learning, and computational modelling — developing tools to quantify uncertainty, guide experimental design, and accelerate crystallisation process development.
I also work on integrating smart machine learning methods for efficient pharmacokinetic modelling, and have explored data-driven high-dimensional time-series monitoring methods for real-time quality assurance in manufacturing.
The past few years I’ve worked on:
I am interested in exploiting advanced numerical and computational techniques to quantify modelling uncertainty, facilitate process design, and support decision-making. Check out my publications:
And check out some of my code: