AI research scientist at Canva Research | Sydney | PhD @UniofOxford | South African 🇿🇦 | @DeepIndaba 🐘 | maker of things
Sydney, Australia
Joined November 2009
- Tweets602
- Following401
- Followers2.1K
- Likes865
Pinned Tweet
Interested in training AI models used by +200M users all around the world? My team at Canva Research is hiring a Senior Research Scientist in Design Generation, based in Sydney! ☀️🔥 Apply here if you're an expert in training LLMs at scale: lifeatcanva.com/en/jobs/6000…
A big milestone for our team and @canva Research today. After many months of hard work, we've released Brush - Canva's very first vision-language model for design editing 🖌️🎨
All those frustrating moments when you've tried to change something specific in an image or design, and the AI tool just failed? Well Brush is here to help. It's trained specifically for editing tasks, learning from examples that show how to make precise edits, step by step.
And what a way to kick things off: our first post introduces Craft, Canva's very own foundation image generation model, which debuted today on DesignArena behind only OpenAI models! An amazing achievement from the teams involved. You can read more here: canva.com/newsroom/news/canv…
Since joining @canva Research, I've been working on a little side project with a few other wonderful folks: our official research website and blog.
Today it went live: canva.com/ai-research 🔥 🎉 🔥
Calling all AI science communicators to submit proposals to the @NeurIPSConf 2026 Education Track! 🔥 We're looking for you to teach the latest technical concepts in AI so that audiences can get the most from the main conference week. Deadline is 4 September AoE ⏰
Calling AI Science Communicators: NeurIPS 2026 Education Track
We're inviting science communicators in AI to submit proposals to teach an emerging technical concept. Concepts can be any technical idea that has underpinned a large set of papers at top-tier AI conferences over the past year, from new architectures to training and alignment techniques to evaluation and reasoning methods. Think the 2026 equivalents of the Transformer, the reparameterisation trick, or LLM judges. You don't need to be the original author of the concept, just able to explain it with real clarity.
Accepted proposals must be paired with original teaching materials (code notebooks, demos or short videos) and will be presented as part of a full day primer in Sydney on 7 December 2026, ahead of the main NeurIPS conference. All materials will be gathered in an open repository, free for the community after the conference.
Submissions close 4 September 2026 AoE (Anywhere on Earth)
Submit via OpenReview: neurips.cc/Conferences/2026/…
Learn more about the track at neurips-education-track.gith…
You don't need to have come up with the concept yourself - just be passionate about teaching it super clearly ⚡ All proposals must be paired with original teaching materials. Selected proposals will be invited to present a talk or a poster in Sydney on 7 December 2026! ⛱️ 🌞
I'm very excited to be co-chairing the NeurIPS Education track this year, alongside @KoenigsteinNoam and long-time @DeepIndaba friend, @ulrichpaquet 🐘 See you in Sydney!
Find My Things, an object recogniser our team built that helps people who are blind find their personal objects, has been featured by @Microsoft's blog on "6 ways AI is making a difference in the world". Check it out along with some other amazing work by @MSFTResearch and others!
AI is enabling change in remarkable ways, from improving health care and education to making life easier for people with disabilities. Read about six ways Microsoft AI is bringing positive change around the world. msft.it/6006UAmBs
A bit late to the party, but happy to share our #NeurIPS2024 paper: arxiv.org/abs/2406.04236! We use a causal tracing methodology to study how multi-modal LLMs like LLaVa and Phi retrieve information required to answer factual visual questions. 🔎❓👩🏼💻
Excellent work led by @BasuSamyadeep in collaboration with @FeiziSoheil, @besanushi, Cecily Morrison, and Martin Grayson 🤓
You can find code for our work here: github.com/microsoft/MLLMInt…
Check out our #EMNLP paper on how we can improve CLIP's visio-linguistic capabilities by adding a loss term which distills this information from a diffusion model (which has stronger visio-linguistic abilities). Great work led by @BasuSamyadeep
This #emnlp2024, we have 3 papers on VLM compositionality, prompt tuning and language model interpretability!
(i) arxiv.org/pdf/2307.09233 answers how we can distill good properties of generative models (e.g., compositionality) to contrastive CLIP like models! [Main]
Final 2 weeks to apply for an ML post-doc role in Equitable AI with our team at @MSFTResearch Cambridge UK! shorturl.at/t8WLU
We're looking for candidates with a PhD in Computer Science, Machine Learning/AI or a related field, a passion for building inclusive AI tech...
...and a strong publication record in any of the following areas: multi-modal generative models, AI fairness/bias, transparency/interpretability, AI ethics, accessibility, data-centric AI, model robustness, OOD generalisation, long-tailed learning, & generative evaluation methods