Stefania Druga
@Stefania_druga
Independent, Former Research Scientist @GoogleDeepMind AI & Multimodal LLM applications/ PhD @UW / alumni @mit @msft @Theteamatx
Last week I introduced LabMind 👩🔬, a real-time AI co-scientist to support researchers with their experiments. You can now watch my talk from @aiDotEngineer conference in San Francisco where I did a live demo of the platform. If you are curious to try LabMind and want to…

Everyone's talking about AI performance on the IMO. Let me highlight 🇨🇦Canadian 11th grader Warren Bei🇨🇦, one of five participants with a *perfect* 42/42. This is his *fifth* (and final) IMO representing Canada, with three golds and two silvers. (➡️ MIT undergrad in the fall)
Pair programming accelerated software development, could "pair science" do the same for research? What if you had an AI lab partner that never sleeps? 🔬🤖 I built LabMind, an 🧪 AI co-scientist that provides 24/7 monitoring, real-time insights, and powerful simulation…

Introducing renaissance vocal percussion baby. Made with Act Two
Kimi K2 is basically DeepSeek V3 but with fewer heads and more experts:
Can an AI model predict perfectly and still have a terrible world model? What would that even mean? Our new ICML paper formalizes these questions One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
Kimi-K2-Instruct is a new open weights model from @Kimi_Moonshot today - it's HUGE (1T parameters, 958.52 GB on Hugging face), maybe the largest open weights model ever? More of my notes here: simonwillison.net/2025/Jul/11/ki…
i think Prompt Engineering is "growing up" in two different ways in 2025. the first is "Context Engineering", which is well covered by @dexhorthy's talk (now >100k views!) but Prompts = Intent + Context and as @sgrove points out the task of encoding intents/goals/principles…
🆕 Structured communication is the bottleneck. Whoever communicates best becomes the most valuable programmer. Today's feature is The New Code, by @sgrove of @OpenAI! As LLMs increasingly do more and more code generation, the need for clear communication with Agents and with…
Thrilled to introduce Foundation Model Self-Play, led by @_aadharna. FMSPs combine the intelligence & code generation of foundation models with the curriculum of self-play & principles of open-endedness to explore diverse strategies in multi-agent games, like the one below 🧵👇
My mom sent me a 30-year-old drawing... That I made for her when I was a kid. Naturally, I animated with Midjourney.. + It's not perfect + But it's awesome + And it captures the innocent chaos + Of my childhood imagination. Yes, my mother is fighting a dragon... With a…
Update: I'm open-sourcing this. My DMs have exploded with investors, pro players, teams, users. The demo has 5,000,000 views across platforms. There is obviously interest. But, I have zero interest in working on it. I'm more interested in working on Tidbit. So -- someone…
used gemini 2.5 pro to build a simple shot counter for myself + give jordan feedback per shot.
If you're a new mom, we're hiring! • Backend: Ruby/Rails, TypeScript/Next • Frontend: React, Next Part-time (10-25 hours a week), work from anywhere, no mandatory meetings, $100-150/hr + equity. If you’re in or near NYC, we’d love for you to come in once a week; if you don’t…
I wish data centers would offer tours to the public and schools could take field trips to them. They are the defining pieces of infrastructure of our generation, but unlike railroads, the grid, or anything else, we never get to see them and experience their scale.
When I get tired and sick from working so hard and optimize for impact not $$ messages like these are my rocket fuel 🙏

SOS kids village in Vietnam building with Lovable 🥹
.@GoogleDeepMind is running a hackathon with a $10,000 Ollama prize. The award is for the best project that utilizes and showcases the capabilities of Gemma 3n running on Ollama. There is a total of $150,000 of prizes to be won! Link 👇👇👇
Such a joy to visit South Korea again! Thank you for the invitation @SeoulMetaWeek ! Will definitely be back soon as it's only a short flight from Tokyo.




I am teaching this online class on building multimodal AI apps next week (nor prior experience needed).Please share with any people who might be interested. Scholarships available for talented and curious students, feel free to DM me :) linkedin.com/posts/drugaste…