Sydney tech founder uses ChatGPT to design cancer vaccine for his dog

When vets say 'months to live', some people accept it

Paul Conyngham has never been the accepting type.

The Sydney-based data scientist adopted Rosie, a Staffordshire bull terrier cross, from a shelter in 2019. Five years later, she developed mast cell cancer, an aggressive skin cancer common in dogs, and neither chemotherapy nor surgery managed to halt its spread. By late 2024, tennis ball-sized tumours covered her leg and vets told Conyngham she had months left.

Conyngham, co-founder of Core Intelligence Technologies and a former director of the Data Science and AI Association of Australia, decided to apply his 17 years of machine learning experience to a problem he knew nothing about: cancer immunology.

Eighteen months later, Rosie received Australia's first personalised mRNA cancer vaccine designed for a dog, and her largest tumour has since shrunk by 75%.

The pipeline: DNA sequencing, AI interpretation, mRNA synthesis

The first step was understanding what made Rosie's cancer cells different from her healthy cells. Conyngham contacted the Ramaciotti Centre for Genomics at UNSW and paid $3,000 to have her tumour DNA sequenced alongside her normal tissue.

This produced gigabytes of raw genetic data showing the mutations specific to Rosie's cancer, though on its own the data remained unactionable without expert interpretation. Conyngham needed to identify which mutations would make good targets for a vaccine.

He turned to ChatGPT as a research assistant. The chatbot explained how personalised cancer vaccines work: scientists identify neoantigens, mutated proteins unique to a specific tumour, and encode them in mRNA. When injected, cells read the mRNA instructions and produce these mutated protein fragments, training the immune system to recognise and attack cancer cells displaying them.

Conyngham then used AlphaFold, DeepMind's protein structure prediction tool, to model how his shortlist of mutated proteins would fold in three dimensions. This helped him identify which mutations would be visible to Rosie's immune system and therefore viable vaccine targets.

After months of analysis, he condensed his work into a half-page formula describing the mRNA sequence for a vaccine targeting Rosie's specific cancer mutations. Associate Professor Martin Smith, director of the Ramaciotti Centre, watched the process unfold and later told reporters that Conyngham's persistence surprised even experienced researchers.

From formula to injection

Conyngham brought his half-page formula to Professor Pall Thordarson, director of UNSW's RNA Institute. Thordarson's team synthesised the mRNA vaccine and encased it in lipid nanoparticles, the same delivery mechanism used in Pfizer's and Moderna's COVID vaccines.

The entire manufacturing process took less than two months from finalised sequence to physical vaccine.

But Conyngham still needed someone to inject it. Australian regulations required ethics approval for experimental animal treatments, and the UNSW researchers who made the vaccine weren't authorised to administer it.

The regulatory process proved more challenging than the science itself, according to Conyngham. He spent three months working through a 100-page ethics approval document, dedicating two hours every night to the paperwork.

The Canine Cancer Alliance, a Seattle-based nonprofit, connected Conyngham with Dr Rachel Allavena at the University of Queensland's School of Veterinary Science, who had existing ethics approval to test immunotherapy in dogs. In December 2025, Conyngham drove 10 hours to Gatton, Queensland, for Rosie's first injection.

Results and limitations

By mid-March 2026, when Conyngham appeared on Australia's Today Show, the tennis ball-sized tumour on Rosie's leg had shrunk by roughly 75%. Her energy levels improved and she regained much of her former personality.

One tumour has not responded to the treatment, and Conyngham is already sequencing it to design a second vaccine targeting different mutations.

Conyngham believes some cancers could potentially shift from terminal diagnoses to manageable chronic conditions using this approach.

Thordarson emphasised that this represents the first personalised cancer vaccine ever designed for a dog, and that the same mRNA technology being tested in Rosie will eventually be applied to human patients. Personalised medicine, he noted, can be both highly effective and produced quickly when the underlying platform exists.

What this actually demonstrates

Rosie's case does not prove that AI has cured cancer, and that framing misses the point entirely.

What Conyngham demonstrated is that the pipeline for personalised cancer treatment, from tumour sequencing to mRNA vaccine, can be compressed from years to months using tools that already exist. DNA sequencing costs have fallen by orders of magnitude over the past decade, protein structure prediction via AlphaFold is free and open source, and mRNA synthesis technology matured during COVID vaccine development.

The AI tools did not replace scientists but made complex biology accessible to a motivated non-expert, who then partnered with qualified researchers to execute the actual science. ChatGPT served as a 24/7 research assistant and tutor, while AlphaFold provided structural predictions that would have required months of laboratory work a decade ago.

Thordarson believes the approach could be democratised in Australia without reliance on foreign pharmaceutical companies, with potential applications beyond cancer extending to neurological diseases.

The uncomfortable question

Conyngham has asked publicly why, if this approach works for a dog, the same technology is not being deployed for human cancer patients at scale.

The answer involves regulatory frameworks designed for different eras, reimbursement systems that struggle with truly personalised treatments, and clinical trial requirements built around standardised interventions. Pharmaceutical companies develop drugs they can patent and sell at scale, and a bespoke vaccine for one patient, human or dog, does not fit that model.

Rosie's case suggests the technology side of the problem is largely solved, with the remaining barriers being institutional rather than scientific.

Dr Nathan Gianneschi, a chemist writing for The Conversation, urged caution, noting that the improvement in Rosie's condition cannot be definitively attributed to the vaccine and that outcomes in one dog cannot be extrapolated to other animals or humans.

Gianneschi is right that this is one dog and not a clinical trial, but that observation is somewhat beside the point. Conyngham was not trying to prove a new cancer treatment to regulatory standards. He was trying to save his dog, and so far, he has.

Whether the systems that govern human medicine can adapt quickly enough to enable similar interventions for people remains the open question, one that extends well beyond any single dog, scientist, or entrepreneur.