AI at Google/Kaggle (inversion). PhD chemical engineer. Too many hobbies.
Joined December 2007
- Tweets 8,203
- Following 2,025
- Followers 9,197
- Likes 9,966
397 Photos and videos
Jun 9
Wink with my left eye. (The right works just fine.)
what’s a completely normal thing other people do that you cannot for the life of you figure out how to do?
172
Earlier today we released local development for Kaggle Benchmarks. 🚀
You can now write, validate and run AI evaluation tasks directly from your preferred dev environment — VSCode, Antigravity, Claude Code, and more.
Go from idea to working eval using natural language with the write-kaggle-benchmarks skill.
9
23
114
16,224
Walter Reade retweeted
Jun 3
Our new Gemma 4 12B model hits a sweet spot between size performance: it can run locally on a laptop, while enabling powerful multi-step reasoning and agentic workflows. Can’t wait to see what the community does with this one!
Jun 3
Celebrating the milestone of a massive 150 million downloads of Gemma 4 with the release of the new Gemma 4 12B model! It's incredibly powerful for such a small model and it’s tiny enough to run locally on a laptop with just 16GB VRAM. Apache 2.0 license - happy building!
240
369
5,030
470,534
Walter Reade retweeted
May 27
It's been awesome working with @xdotli @nick_kango @ivanleomk @WalterReade @ashley_oldacre to bring this hackathon to Kaggle & the community!
How well can you write skills that improve fixed agent harness performance on a variety of tasks? Submit skills & writeup findings. 👇
When an AI agent succeeds, was it the model or the skill it was given?
Launching today with @xdotli and @benchflow_ai — the BenchFlow AI Agent Skills Community Hackathon. Build skills that lift agent capability without crossing safety boundaries.
3
16
1,872
Walter Reade retweeted
May 23
We added a model filter to Kaggle Benchmarks!
Check it out to see the models we support. Anyone can eval these models at no cost on our platform. :)
Plus, of course explore results for models you're interested in!
1
6
28
3,116
Kaggle は昇進の役にたつ
“入社3年目でスペシャリストグレードに昇格しており、20代で同グレードに到達している社員はいないという。スペシャリストグレードへの昇格には、社内での業務実績に加え、社外でも通用する専門性が求められる。Kaggle Masterの称号や技術力の高さが評価された”
［ITmedia ビジネスオンライン］Kaggleトップ1%のAI人材、なぜNTT東日本に？ 若手データサイエンティストの活躍 itmedia.co.jp/business/artic…
9
160
47,773
May 22
Technical writing these days:
- Step 1: Ask AI to write the paper
- Step 2: Ask AI to review the paper for signs that it was written by AI
- Step 3: Ask AI to suggest less AI-sounding phrases
- Step 4: (Optional) Read the draft before submitting
4
4
33
2,169
May 20
I should note . . . the name of this competition is :chef's-kiss:
I'd join if I wasn't working on 19 other things. 😅
May 20
Help the Natural History Museum of Denmark automate label transcription on a Danish Dung Beetle collection.
MuseumSCAT: Specimen Collection Annotation Task@CVNH ECCV26
kaggle.com/competitions/muse…
1
1
1
405
May 20
Help the Natural History Museum of Denmark automate label transcription on a Danish Dung Beetle collection.
MuseumSCAT: Specimen Collection Annotation Task@CVNH ECCV26
kaggle.com/competitions/muse…
2
4
775
May 19
I jumped into machine learning ~14 years ago. It was an exciting and fun time. The last few years, though, have felt like a grind trying to keep up with everything. Agentic coding has made things fun again.
18
934
May 19
Rookie numbers.
May 18
Don't let anybody talk you out of spending $50,000 on anthropic api credits.
That $100,000 will be the best $250,000 you ever spent.
1
2
961
May 17
I consent to having AI trained on all of the papers I've published over the years. (That, or Elsevier can start paying authors royalties.)
Science publishing giant Elsevier has joined the dozens of firms and individuals suing artificial intelligence companies over their alleged use of copyrighted works in training AI models
go.nature.com/4uUldNO
3
11
2,004
Walter Reade retweeted
May 15
we're launching kaggle benchmarks local development capabilities v soon (i.e., you can build your benchmarks in your own IDE with agents!!)
lmk if you want to join our early access program:)
3
1
13
1,275
Apr 25
Who else feels this way: If a task uses to take 10 hours to complete, and now AI can do the first 9 hours of work in 10 minutes, that last hour of non-AI work feels 10x harder.
Anyone feeling this?
5
1
13
1,090
Apr 23
This looks like a fun tabular community competition.
kaggle.com/competitions/info…
1
3
25
2,425
Walter Reade retweeted
Apr 19
This is a massive bug, not a feature. Claude shouldn't need positive affirmations and 'good boy' prompts to stay sharp.
Users are paying a lot of money to Anthropic and now they’re telling people they need to babysit Claude’s emulated emotions because they causally shape outputs????
5
5
195
3,749
Walter Reade retweeted
Apr 19
I'm afraid to enter.
There wold be no return.
I will enter.
3
2
17
1,458