Postdoctoral Researcher & AI Engineer
Machine Learning & Representation Learning for Chemistry & Materials Sciences
Eduardo Alberto Aguilar-Bejarano
Nottingham, UK (visiting researcher, MIT, Cambridge, MA — Jun–Aug 2026)
ed.aguilar.bejarano@gmail.com
Chemist and machine-learning researcher developing data-driven approaches to fundamental problems in chemistry, materials science, and engineering. I am particularly interested in incorporating domain knowledge, from physical chemistry to biology, into ML models, and in using interpretable models to turn predictions into testable structure–property hypotheses. My work spans asymmetric and organometallic catalysis, polymers, and lipid nanoparticles, developed in close collaboration with experimental groups and validated prospectively on newly synthesized systems. I also build AI workflows and open-source tools that meet the practical needs of experimental scientists, accelerating experimentation while reducing research costs and environmental impact. I believe AI should help us solve cutting-edge problems faster without sacrificing a fundamental understanding of the phenomena behind them.
Based at the Koch Institute for Integrative Cancer Research (Cambridge, MA). Developed interpretable computational models for the design and prioritisation of lipid nanoparticles (LNPs) for more efficient drug delivery, extending prior LNP composition-analysis work toward model-guided exploration of lipid chemical space. Continuation of the Nottingham–MIT LNP collaboration with Prof. Daniel Anderson's lab.
On the EPSRC Large Grant: Designing bio-instructive materials for translation-ready medical devices (Advisor: Dr. Grazziela Figueredo). First author and architect of Helix 1.0 (CI/CD-managed framework for reproducible, interpretable ML on tabular scientific data; applied to biomarker discovery of macrophage cell types from high-dimensional metabolomics data). Lead developer of PolyNet (end-to-end Python/Streamlit library for structure-based polymer property prediction) and of LCNet / LNPQuantification (the neural-network model behind the first quantitative Q-Cryo-OrbiSIMS method for LNP characterisation, with MIT's Anderson Lab — curated the 337-spot calibration dataset; predicts the four canonical LNP lipid components to ~14 mol% MAE). Lead modeller on an antibiofilm polymer-discovery programme (ensemble GNNs over ~500 polymer structures against high-throughput biofilm assays for P. aeruginosa, S. aureus, and uropathogenic E. coli). Also delivered in-house data-science and software training and promoted open-source, reproducible research across an interdisciplinary EPSRC team.
Doctoral thesis studying chemical representation across three systems of increasing difficulty: (1) transition-metal complexes — interpretable GNN pipeline for enantioselective catalyst/ligand optimisation (→ HCat-GNet, iScience; AsymBench, J. Cheminform.; data curation, Molecules); (2) transition-metal carbides — new 3D-masking explainability for crystal-convolution GNNs and machine-learnt interatomic potentials, applied to molybdenum carbides (→ PCCP); (3) polymers — ensemble-GNN methodology for virtual screening and multi-property optimisation of polymer biomaterials (→ PolyNet).
Cheminformatic modelling for in-silico prediction of physicochemical profiles and rational design of bioactive compounds (Advisor: Prof. William Zamora-Ramírez). Built a novel lipophilicity descriptor from toluene/water partition-coefficient prediction (MLR, SVM, random forest, gradient boosting); developed the first multivariable model for tautomer equilibrium constants in water, extended toward organic solvents; and characterised antimicrobial-peptide structures using QM continuum-solvation calculations.
First author unless noted. Full list with graphical abstracts on the Publications page.
Lead/first-author role unless noted. Most publications ship with an accompanying public code repository.
Reviewer for the following journals. Click a linked journal to view the reviewer certificate.
Designed and delivered in-house Python and ML training to interdisciplinary research teams.
Co-supervised three MSc students in 2024 with Prof. Simon Woodward, and one MSc student (Veera Pedireddi) in 2023 with Dr. Grazziela Figueredo and Dr. Kristaps Ermanis, on deep-learning projects in the chemical sciences.
Assisted across seven laboratory courses: Physical Chemistry I & II, Analytical Chemistry I, Organic Chemistry I, General Organic Chemistry I, Chemical Experimentation II, and General & Qualitative Chemistry.
Teaching assistant for Prof. William Zamora's Cheminformatics course (programming, data visualisation, ML algorithms, cheminformatic tools).
