EXPERIMENTAL BIOLOGY · QUANTITATIVE IMAGING · EVIDENCE-GROUNDED AI
I connect experimental cell biology, quantitative imaging and computational modelling to investigate biological organisation, predict cellular behaviour and build traceable scientific platforms.
My work spans biological feature design, prospective validation, multimodal evidence integration and human-guided interpretation—keeping models accountable to biological evidence.
Research
Protein architecture, cellular organisation and perturbation evidence across the TRIM family.
Explore →Causal mechanisms resolved through structure, interactions, kinetics and perturbation.
Explore →Chromatin dynamics, perturbation responses and nuclear state-space analysis.
Explore →Mesoscale organisation measured across concentration, variants and cellular context.
Explore →Cross-scale analysis of biological organisation, function, robustness, plasticity and evolvability.
Explore →Variant evidence, molecular consequence, cellular state, biomarkers and therapeutic hypotheses across distinct validation gates.
Explore →Methods & systems
Oscillations, pulses, synchronisation and perturbation–recovery dynamics across biological time series.
Explore →Method selection across molecular, cellular, tissue and organismal scales, with controls and inference boundaries.
Explore →Signals, feedback, memory, noise and perturbation across biological information systems.
Explore →Target validation, biomarkers, modality, uncertainty and next-experiment selection.
Explore →Biological state, transitions, memory, feedback, perturbation and recovery across interacting scales.
Explore →Claims, figures, uncertainty, collaboration and release review with provenance preserved.
Explore →
▶ Watch video
Sequence, domains and interactions — the modular architecture TRIMCode connects to cellular organisation and, downstream, to predictive modelling.
▶ Watch video
A proposed computer-vision framework converting microscopy into analysis-ready masks — reference annotations, predictions and experiments kept explicitly distinct.
▶ Watch video
Sequence, predicted structure and disorder combined into a proteome-scale prediction of condensate-forming proteins, validated against cellular evidence. Developed with the Tóth-Petróczy lab, MPI-CBG.
▶ Watch video
The open question connecting sequence-level features to condensate-forming behaviour — the boundary this Bio × AI work, with the Tóth-Petróczy lab at MPI-CBG, is testing against cellular evidence.
Models, validation and ensembles through inspectable prediction and failure.
Explore →Relational querying through a local, inspectable evidence trail.
Explore →Reproducible animated flow visualisation with topology-stable transitions.
Fifty interactive experiments in prediction, representation and diagnosis.
Explore →Representations, optimisation, memory and attention through manipulable visual experiments.
Explore →Compute, storage, orchestration and observability from local systems to production infrastructure.
Explore →RAG, tools, state, approval and evidence-grounded orchestration with explicit autonomy boundaries.
Explore →Retrieval, ranking, constraints, evaluation and safe agent-assisted recommendation.
Explore →Public evidence, provenance and executable examples across the Bio × AI learning system.
Explore →Scientific visual reasoning through distributions, uncertainty, time, networks and meaningful motion.
Explore →PCA, clustering, nonlinear maps and anomaly detection without treating structure as ground truth.
Explore →Agents, value, policy, evaluation and bounded sequential decision-making.
Explore →Biological text representation, retrieval and citation-grounded reasoning with provenance.
Explore →Temporal diagnostics, honest baselines, forecasting uncertainty and decision boundaries.
Explore →Prediction from biological sequence while keeping scores and rankings inside their evidence boundary.
Explore →Detect when a deployed model has stopped representing the system it was built for.
Explore →Leakage, split design and validation scores that retain their intended meaning.
Explore →Randomisation, confounding, peeking and the limits of statistical inference.
Explore →Probabilities, intervals and confidence without treating agreement as ground truth.
Explore →Cross-validation, leakage-safe tuning, model comparison and test-once evaluation.
Open Model Selection Playground →Emerging initiatives
Causal perturbation models, mechanistic state transitions and digital-twin hypotheses for how biological systems respond, recover and change.
Assistive systems for evidence synthesis, experimental planning and reproducible analysis, with bounded recommenders for candidate ranking and abstention.
Knowledge graphs, ontologies and provenance-aware retrieval across literature, mechanisms and biological evidence.
Multimodal, spatial and single-cell systems for organising biological measurements, relationships and cross-scale data.
Evidence and utility structures connecting disease biology, biomarkers, therapeutic hypotheses and programme-level decisions.
Inspection-first software for bioengineering design–build–test–learn cycles, experimental methods and reproducible biological analysis.
Multidisciplinary life sciences
Designing, testing and translating biological mechanisms into engineered living systems and useful biotechnology.
Revealing structure, dynamics, energy and interaction through physical principles, chemical reasoning and multiscale measurement.
Turning quantitative and imaging-derived observations into models, software, AI systems and engineered tools that generate testable scientific decisions.
Highlights
For resilient environments and a livable future.
Open highlight → Sustainable prosperityBiological innovation for sustainable prosperity.
Open highlight → Dignity & accessScience and systems in service of human dignity.
Open highlight → Personal archive · 2017The thread connecting these directions, preserved as first published.
Open the full page →