rasgaard.com - Profile | ThreadSky | a Reddit-style client for Bluesky

Here we go! The Copenhagen NLP Symposium is starting up with a welcome from @delliott.bsky.social. People are attending from Aarhus, Aalborg, Copenhagen, and outside Denmark, from both academia and industry.

submitted 1 day ago • 2 comments

CPH NLP Symposium 🤗🔥 cphnlp.github.io

submitted 1 day ago • 0 comments

I feel like this shouldn't be as groundbreaking as it appears to me but being able to simply write uv run --env-file .env <script.py> to manage which env-file is being used for a particular script execution is so nice docs.astral.sh/uv/concepts/...

submitted 4 days ago • 0 comments

20 år siden i dag (unpossible!)

submitted 8 days ago • 1 comment

Recently read Zen and the Art of Motorcycle Maintenance for the first time. It was a bit of a tough and slow read. Had to consult a guide/explainer quite often but wow was it worth it #booksky

submitted 10 days ago • 0 comments

Do we still need large language models in the era of agentic AI? Agentic AI systems are driving a wave of applications where language models carry out a limited set of specialized tasks repeatedly, with minimal variation. "Small Language Models are the Future of Agentic AI"

submitted 16 days ago • 3 comments

Here's a great little tutorial on getting started with my LLM tool for running local models! One of the best ways you can contribute to any open source project is to publish unofficial guides like this one

submitted 19 days ago • 2 comments

I have soft-launched a newsletter of sorts. It's just a few links to things that I have liked enough to note down throughout the week. I like the idea of having a big archive of things that I like and why I like them :) rasgaard.com/newsletter/

submitted 19 days ago • 0 comments

I love everything about this mind-boggling analysis of the latest season of #therehearsal on HBO. It makes me appreciate Nathan Fielder's work even more which I didn't think was possible. www.reddit.com/r/TheRehears...

submitted 20 days ago • 0 comments

submitted 22 days ago • 24 comments

My first podcast appearance is live :) I was a bit nervous but it was a lot of fun! We talked about running machine learning models locally/client-side in the browser.

submitted 22 days ago • 0 comments

I have been thinking about this as well. You shouldn't be afraid of creating bespoke tooling for your AI annotations and evals. The time spend finding a tool that exactly matches your workflow is often better spent implementing a tailor-made solution.

submitted 22 days ago • 0 comments

I have written a barebones guide to getting started with transformers.js :) I hope that data scientists and web devs can meet somewhere in the middle to take proper advantage of client-side ML inference in the browser #mlsky #ai #dkai rasgaard.com/posts/gettin...

submitted 23 days ago • 0 comments

Haven't ordered new books in a good while because I have been using the library so much but now that I got a few gift cards for my birthday I checked out these :) #booksky

submitted 23 days ago • 0 comments

how else can you describe the rehearsal other than just “pure art”

submitted 24 days ago • 89 comments

Wow: Bluesky drives almost the same traffic to my blog (The Pragmatic Engineer Blog) as Twitter/X Data from last month. Just incredible that a site with ~33 million users and growing fast (Bluesky) punches as much as one that is supposedly 20x the size (Twitter/X is estimated around 600M users).

submitted 25 days ago • 17 comments

All the ways I want the AI debate to be better by Andy Masley https://andymasley.substack.com/p/all-the-ways-i-want-the-ai-debate #AI #tribes

submitted 26 days ago • 1 comment

My latest little hobby project: Fine-tuning a tiny 40M Whisper model specifically for Danish. Benchmarks look great! But how is practical performance? Not amazing..:) rasgaard.com/posts/whispe... #dkai #ai #mlsky

submitted 27 days ago • 2 comments

Velkommen til endnu flere seje tech-profiler: @peterdalsgaard.bsky.social @jonasfrich.bsky.social @julielykkes.bsky.social @olifent.bsky.social @petersvarre.bsky.social @sunelehmann.com @adler-nissen.bsky.social Sig til, hvis du kender nogen, der skal på listen! go.bsky.app/RDnpz5p 1/2

submitted 27 days ago • 1 comment

I have been working on bootstrapping AI evals before product launch and came up with something like this. Using QA to inform "problematic on purpose" synthetic data that can validate LLM-as-a-Judge. Would love to chat with people who have experience with this :) #mlsky #buildinpublic

submitted 29 days ago • 0 comments

I usually don't really like off-the-shelf AI evaluation tools. Like, what do you even mean when you measure "groundedness"? Using AI Toolkit for VS Code for LLM-as-a-Judge alignment evals has been the first time I have had an okay experience with a tool like this.

submitted 30 days ago • 0 comments

EuroSpeech: Massive Multilingual Parliamentary Speech Corpus - 78,100+ hours across 22 European languages - 50,500+ hours of quality-filtered data (CER < 20%) - Robust alignment algorithm for non-verbatim texts - Dramatically expands resources for 19+ languages huggingface.co/datasets/dis...

submitted 31 days ago • 0 comments

What a time to be alive

submitted 31 days ago • 0 comments

Gemma 3n E2B available through Edge Gallery: github.com/google-ai-ed... Running on my Nothing Phone 1 here. #MLSky

submitted 32 days ago • 1 comment

Driving AI hos @ida.dk er i dag. Glæder mig især til at høre om: - Region Hs erfaring med AI til håndtering af hudkræft - Generativ AIs miljøpåvirkning af folk fra ddsc.io - Vision Transformers fra Capacit som havde et maks-nørdet oplæg sidste år om DSPy. #dkai #dktech

submitted 32 days ago • 0 comments

Vil gerne blive bedre til at dele hvad jeg har gang i. Også selvom det måske er en smule upoleret. Her er sidste nyt: Whisper-large-v3-turbo som web app, 100% lokalt så din data aldrig forlader din computer. Virker bedst i Chrome! (pga. webgpu support) rasgaard.com/p/transcribe/

submitted 33 days ago • 0 comments

Har leget med en idé om annotering som del af servicen: - Simpel Whisper-app der kører lokalt -> gratis med Transformers.js - Fokusér på UI/UX så det er lækkert og nemt at bruge (nok det sværeste) - Gør det muligt at rette fejl i teksterne - Optional: Indsend rettelser + lyd til database

submitted 34 days ago • 1 comment

Low Tech AI!

submitted 36 days ago • 2 comments

Happily doing propaganda for small models and the new generation of open source AI startups in a very nice company (Pleias, Moondream, Prime Intellect). globalventuring.com/corporate/eu...

submitted 37 days ago • 0 comments

AttentionInfluence: for pretraining data selection Good data matters, but how do you find it? This paper uses the attention heads from existing models to calculate & rank how valuable the data will be during training Mask out critical heads and calculate the loss arxiv.org/abs/2505.07293

submitted 38 days ago • 4 comments

[28/30] 86 Likes, 6 Comments, 1 Posts 2504.20752, cs․CL | cs․AI | cs․LG, 07 May 2025 🆕Grokking in the Wild: Data Augmentation for Real-World Multi-Hop Reasoning with Transformers Roman Abramov, Felix Steinbauer, Gjergji Kasneci

submitted 39 days ago • 1 comment

Danmark har et Microsoft-problem.

submitted 40 days ago • 0 comments

By the time the EU will make up its mind on LLM transparency regulation, pretraining in its current shape won’t exist anymore.

submitted 43 days ago • 2 comments

Er virkelig taknemmelig for at DHH bliver inviteret til at dele ud af sin erfaring med open source når det handler om Danmarks afhængighed til Amerikansk tech #dktech #dkpol Hans tale her er meget anbefalelsesværdig! www.linkedin.com/posts/david-...

submitted 43 days ago • 0 comments

It is very easy to make mistakes when creating evals for your AI product. @sh-reya.bsky.social and I run through the most common errors in this talk. 35% discount code to our upcoming course in the video notes youtu.be/GL0XhAj5LPE?...

submitted 44 days ago • 0 comments

The “Paper Skygest” is a total validation of the bluesky thesis. Anyone can build a useful, tunable feed. It’s a bit sparse right now but it’ll be amazing once it takes off fully.

submitted 45 days ago • 2 comments

Er imponeret over hvor godt Qwen3-0.6B kører med Transformers.js. Det hele er lokalt i browseren(!) Det her er et minimalt hjemmebrygget eksempel for at forstå lidt af hvad der sker under motorhjelmen. @xenova.bsky.social har lavet en bedre demo :) Prøv den her: huggingface.co/spaces/webml...

submitted 45 days ago • 1 comment

PyData Copenhagen @ Google Denmark's offices 🐍🔥

submitted 46 days ago • 0 comments

To Bluesky feeds som jeg har været ret glad for på det sidste: ML Ranked Feed: Posts om machine learning & AI ranked på samme måde som HackerNews bsky.app/profile/did:... Build in Public: Folk der laver apps, services, etc. og poster om det. Hyggeligt at følge med. bsky.app/profile/did:...

submitted 46 days ago • 2 comments

Jeg er ikke den store smalltalker. Jeg har det bedre med samtaler om noget, der er personligt, vigtigt, interessant eller underholdende. Men det er for dårligt, siger de ekstroverte, vi skal smalltalke mere. Så jeg satte mig for at blive blive bedre til det. www.zetland.dk/historie/sRn...

submitted 51 days ago • 1 comment

My kids and I vibe-coded a simple chore tracking app over the weekend 🙌 It was really fun. Now we'll see whether it actually works 😅 We're going to try it for a week and see how it goes.

submitted 47 days ago • 2 comments

Træner Whisper-tiny på Coral-datasættet. Er spændt på hvor meget man kan få ud af en så lille (38M) model. Som bonus er det fuldkommen gratis: Det kan lade sig gøre på free-tier GPU compute på Kaggle :)

submitted 47 days ago • 1 comment