Soonson Kwon

Global AI Developer Programs Lead @Google / Founder of / Change the world for creative minds / Opinions are mine

[Aritra @ariG23498]

Wrote a short essay on Pallas and how to think about it! Criticisms and feedback are welcome.

I am reading about kernels and hardware for quite some time now, it was time to put them into "blog brain dumps". If I have enough encouragement, I might as well do a series on it (what do you all think?)

[Martin Andrews @mdda123]

I re-built the tiny-diffusion repo (which has 10MM param GPT and diffusion-LLM models for ~Shakespeare text) in JAX.

And to show off more recent methods, I added flow-matching generation too.

@GoogleDevExpert #TPUSprint

[Sayak Paul @RisingSayak]

Sharing on behalf of a friend:

"""
New compute & funding opportunity for researchers!

We're launching the 2026 Google TPU Research & Education Awards.

We want to support researchers pushing the boundaries of ML systems, performance, and efficiency, and those working in high-impact applied science.

What's included:
- Compute: Free access to the latest generations of TPUs.
- Funding: An unrestricted gift to support or partially support a grad student.
- Credits: Google Cloud credits to support your work.

If you or a colleague are exploring what is possible with TPUs, we want to hear from you.

Read the RFP and apply here: https://t.co/mu6GvkKx3D

If you have questions, email [email protected]
"""

[Aritra @ariG23498]

I just completed the youtube series on JAX AI Stack by @robert_crowe.

The idea:
- why is jax important
- how is flax nnx really close to the torch design
- using optax and orbax for optim and checkpointing

https://t.co/UnMk5XlEGh

[Soonson Kwon @ksoonson]

@JeffDean I had a lot of fun creating this. Thank you for giving me permission to make it, even if the Lego face doesn't look exactly like you! :-)

[Taha ⵣ @mlnomadpy]

tell me you are a research nerd without telling me you are a research nerd

me: just got lego of @JeffDean and @sanjayghemawat with there legendary paper map reduce

[Soonson Kwon @ksoonson]

With the AI GDEs at Google Cloud Next 25! It’s always a great feeling to meet with them together.

[Kaggle @kaggle]

We’re excited to announce that Kaggle’s 5-Day Generative AI Intensive Course with Google is now available as a self-paced Learn Guide!

https://t.co/9U0WYQvZXx

[Jeff Dean @JeffDean]

A very cool use of Gemini models to help summarize (and turn into podcasts) the dizzying array of awesome research papers, for those of you attending or interested in @NeurIPSConf in Vancouver this week!

Nice work, @algo_diver!

[Soonson Kwon @ksoonson]

@JeffDean I very much enjoyed your session although I was indeed a bit distracted by your costume. :-)

[Soonson Kwon @ksoonson]

@dsmcl Thank you for all of your support too! I was so happy that I walked for 2 hours after the graduation event with reflecting the good memories.

[Soonson Kwon @ksoonson]

+1,000 developers graduated the Korea Google ML Bootcamp for the past 5 years. Coursera Deep Learning Specialization and Kaggle competitions are part of the intensive curriculum.

This is a big milestone. I am really happy and proud of being part of the team.

[Google Korea @googlekorea]

지난 10월 18일, 구글 머신러닝 부트캠프 2024가 5기 참가자들의 졸업식을 끝으로 3개월간의 여정을 마무리했습니다! 🎓

489명의 미래 머신러닝 엔지니어들과 함께한 이번 5기 부트캠프에서는 새롭게 추가된 💡 ‘젬마 스프린트’ 과정을 통해 단기간에 프로젝트 아이디어를 제안하고 앱, 코드, 모델 등의 결과물을 만들어 내며 언어 모델에 대한 역량을 강화할 수 있었답니다.

더 자세한 내용은 구글코리아 블로그에서 확인해보시기 바랍니다!🔗✨
👉 https://t.co/lzamlEH6N0

[Soonson Kwon @ksoonson]

@kanpuriyanawab @fchollet @Google @JmlrOrg @penstrokes75 @mattdangerw @jbischof1 @divyasheess @luke_wood_ml @martin_gorner Congratulations and happy to see this!

[Soonson Kwon @ksoonson]

Among the many many exciting conversations with the Indian ML GDEs during the IO Connect Bangalore 2024!

[Soonson Kwon @ksoonson]

It was great to talk with you too!

[Soonson Kwon @ksoonson]

구글 머신러닝 부트캠프 참가신청 시한이 6월 5일 수요일 오후 5시까지로 이제 일주일이 채 남지 않았습니다. 올 여름에 열심히 공부해 보실 분들에게 정말 좋은 기회입니다. 부디 많이 지원해 주시고 홍보도 많이 해 주세요!
https://t.co/7f9h4Xf9Yi

[Soonson Kwon @ksoonson]

Thank you. I am glad to see this happening and I wish it could be an opportunity for the developers/students/researchers there to go deeper into the fundamentals of AI/ML.

[AyeHninn @AyeHninn]

As a ML GDE, I am proud to organize first-ever ML Paper Reading club dedicated for Myanmar ML Community.
Notably, 90 percent of the presenters and 80 percent of the participants are women actively working in the machine learning industry.

https://t.co/3prnmMtho0

[Soonson Kwon @ksoonson]

At Google IO with ML GDEs. :-) @margaretmz @_davidcardozo @perfinion @jigyasa_grover @leslysandra

[Aakash Kumar Nain @A_K_Nain]

Something that I worked on over the weekend: A port of Mistral 7B in Keras with JAX backend. A few things left in the TODO list, but coming out good so far.

PS: I was blocked by an issues while porting Mistral in Equinox, but will get back to it as well

[Colaboratory @GoogleColab]

Check out this Beginner's Guide to Colab produced by one of Google's Developer Experts for Machine Learning, @Sam_Witteveen. Thanks for the video Sam!

https://t.co/8ObKvJ2raF