Research
That is three claims, not one. Something can be perfectly readable and still not worth believing. Reading changes the thing read: sequencing consumes the sample, biodata requires surveillance of a person, a confidence score hides local uncertainty under a global mean. And who is permitted to read at all is a separate question again, which I have been asking since I wrote in 2012 that this shouldn't be limited to men in labcoats.
The question of what biological matter can be made to carry, and who controls it, runs through the work from the 2011 dissertation on the microbiome and the 2012 rotifer farm, through the currency and index artefacts of 2012 to 2019, into the DNA-storage studies from 2020 and the molecular-trust work of 2026.
How a living thing is made legible is not neutral. It changes what people take it to be, and what they think it warrants.
The moment worth studying is the contradiction between what a material shows and what its record claims, because that is where someone has to decide whether to trust the record or the thing.
Some of the evidence for that. In my doctoral study, identical microbial growth shown as live or as a recording produced significantly higher positive affect, immersion, competence and flow in the group that knew it was alive. In a CHI 2022 study, participants told their data could be stored in DNA raised gatekeeping, integrity and memory-alteration fears without being prompted. And in the computational pilots of 2026, global confidence scores did not separate the candidates that mattered: a model score of 96 sitting on top of nine residues below 70 is the reason a triage table has to exist at all.
As information is increasingly written into and read out of living matter and models of it, what does it take to make that matter legible and believable, what does the reading do to it, and who is able to read? Three strands.
DNA data storage as the cleanest case of a record carried by matter. Built on HCII 2020, Mindtrek 2021, CHI 2022, and the Trusted Molecular Memory report with Tom Heinis and Larissa Pschetz. The instrument is a typed, machine-readable trust state that carries identifiers, an evidence summary, inferred and recorded events with their probabilities, and severity-graded flags where the sample and its paperwork disagree.
Whether a model's confidence is grounds for acting. Run on a national compute allocation across four computational projects in 2026, roughly 1,700 predicted structures. None of them synthesised. This is a computational and design studio that commissions wet validation from partners, and I would rather say that here than have anyone find out later.
Who is able to read, and what it takes to make molecular work legible to people who will never enter a lab. Molecular sketching, an accessibility audit of AI-driven protein design workflows, and pedagogy for non-experts.
Sole applicant on a UKRI AIRR Gateway award: 10,000 GPU-hours on Isambard-AI, the UK's national AI research resource. Ref AIRR-GW–2025-07-17, awarded December 2025. The award letter values the allocation at £45,000 and states that it is compute time only, with no funding included. Four computational projects ran on it in 2026 using ColabFold and AlphaFold2-PTM in containers under Slurm, with Python analysis workflows: a pre-synthesis readiness benchmark, a candidate-triage and structural-comparison study run as a documented sprint in May 2026, a stress-test atlas for safe protein screening, and a ferritin pilot on shared folds with different functions.
A technical report with Tom Heinis (Imperial) and Larissa Pschetz (Edinburgh), for ARIA's Trust Everything Everywhere pre-programme. A design-led taxonomy of molecular trust, a molecular chain-of-custody record that travels with a sample, and three stress-test scenarios. Its conclusion is that trust is procedural rather than intrinsic, and that interpretation is a structural part of how trust gets produced rather than an afterthought.
A three-day workshop, July 2025, on what structure prediction does to design practice.
With Jiwei Zhou, on accessibility barriers in AI-driven protein biodesign workflows. Mindtrek 2025.
In research through design the artefacts are the findings, so most of what I make is built to be used and evaluated rather than illustrated.
The molecular trust state, a typed object carrying evidence, inferred events and inconsistency flags.
The molecular chain of custody, a provenance record that travels with a sample.
The build, hold or kill triage tables from the 2026 computational work.
The molecular sketching kit.
The surfacing taxonomy from CHI 2023, which orders six ways of making microbial activity visible by how far each one intervenes in the organism.
2022 to 2024
Two years as a postdoctoral researcher in the Materials Experience Lab at TU Delft with Elvin Karana, on the EU-funded ELM project, asking how designers and users relate to a material that is alive.
Surfacing livingness in microbial displays. A design taxonomy for HCI. CHI 2023.
Cyano-chromic interface. On aligning human and microbial time. DIS 2023, honourable mention.
Living Therapeutic Skin. On-skin living artefacts. Honourable mention at WAD 2024, published by Springer in 2026.
Designing living artefacts for multispecies interactions. International Journal of Design, 2024.
2015 to 2021
The doctoral work at Queen Mary University of London put living microbes inside computer games and watched what happened to the players. Slow games, mouldy glitches, E. coli as a co-player. The thesis is Effects of Microbial Integration on Player Experiences of Hybrid Biological Digital Computer Games, 2020.
A New Mould Rush and MouldCraft. Bio-digital games driven by living fungi and bacteria. Shown at Now Play This (Somerset House), SynbiTECH, and Biodesign Here Now during London Design Festival. FDG 2018; FDG 2020, Exceptional Paper Honourable Mention.
Navigating imaginaries of DNA-based digital data storage. On what people think they are being promised when told their data can live in DNA. CHI 2022.
DNA as digital data storage. On what the medium asks of the people who would use it. HCII 2020.
Microbe-HCI. I led the CHI 2021 special interest group that named the sub-field, and edited a Frontiers in Computer Science special topic on it.
2012 to 2019
Design Interactions work asking what synthetic biology would do to money, food and the home. A banknote has no properties that make it worth anything. Its value is entirely a matter of what people are willing to warrant, which turns out to be the same question I am now asking about a confidence score, run on a substrate where it has nowhere to hide.
Microbial Money. Part of Blueprints for the Unknown, the RCA's project within the EU Studiolab programme. Shown at Science Gallery Dublin, Z33 Hasselt, V2_ Rotterdam and Lafayette College.
Peck As You Go. A biological payment system, and a proposal about dual technologies as much as about money. Winner, Future of Money Design Award, Consult Hyperion, 2015.
Gene Hunting Device. On who owns horizontal gene transfer.
Reviving Drachma. Conceived in response to the Athens General Index.
Rotifer Farm and Biohack the Economy. Exhibited across ten countries between 2012 and 2021, including the V&A, the Saint-Étienne Biennale, Helsinki Design Week and Festspielhaus Hellerau.
Featured in William Myers, Bio-Art: Altered Realities (Thames & Hudson, 2015).
Five episodes of wet-lab work between 2003 and 2018 sit underneath all of it. Protein chemistry at Roche in Basel, 2003 to 2004, on PDHK2 folding and purification and on ClpB cloning and expression. Rotifer molecular biology at OIST in Okinawa, 2011. A biohacking notebook kept through the RCA, 2011 to 2012. A year running Jake Baum's malaria lab at Imperial College, 2014 to 2015, which is where I learned biosafety, procurement and how to keep other people's experiments alive. Cyanobacteria work in the Hanke lab at Queen Mary, 2017 to 2018.
Sixteen organisms handled, thirteen plasmids of which I built eight, protein chromatography across four column chemistries, work at BSL-1 and BSL-2 in a CL3-derogated facility. I don't run a lab now, and I don't claim to. I still think most design conversations about biology would improve if more of the people in them had pipetted.
Three questions I'd want a doctoral student to take on.
Building the inference layer that reads lifecycle events out of molecular signal and scores them against declared history, then studying what laboratory staff actually do when it flags a contradiction.
Reading data out of DNA changes it: sequencing consumes molecules, amplification introduces error, regeneration breaks material continuity. How do practitioners decide what a read-back is worth, and what would a custody record that travels with the sample have to carry?
Model confidence is treated as a go signal. How are candidate-selection decisions actually made in small labs, and does surfacing local uncertainty change what people pick?
New work usually goes out in the newsletter first.