Darek Kłeczek

Machine Learning, Kaggle and occasional pictures from Poland. LLM/AI Research at Snowflake. 4x Kaggle Grandmaster. Personal stuff only.

[Darek Kłeczek @dk21]

Our Training & Finetuning LLMs course is now LIVE 🚀🔥🎉

Brought to you by @wandb and @MosaicML ❤️

Join us for free: https://t.co/Uu45r5L1KO

Get ready to learn from the best in the industry, including experts from @Meta and @Microsoft 😍

🚀 Master Parameter-efficient Training Techniques
🔍 Dive Deep into LLM Evaluation
📚 Explore Datasets for LLM Training
💻 Discover Essential Hardware Requirements

This course isn't for beginners - be prepared for practical insights and the opportunity to train your own LLMs! 💪

[Darek Kłeczek @dk21]

@HamelHusain @isaac_flath Woah, congrats Hamel!!!

[Darek Kłeczek @dk21]

@FishAudio DRAMA

[ML in PL @MLinPL]

We're excited to welcome Snowflake, one of our Gold sponsors, who will be taking the stage to talk about Evolving AI Functions: From Unstructured Data to Agentic Intelligence.

AI Functions bring generative AI directly into SQL, enabling scalable semantic processing of unstructured and multimodal enterprise data. But making these functions truly useful requires solving a deeper research problem: how do we continuously improve their quality, cost, and reliability as models, workloads, and user needs evolve? In this talk, Darek will share research from Snowflake on evaluation and benchmarking, as well as automated function evolution, model routing and cascades, and agentic optimization.

The talk will be presented by @dk21, Staff Research Scientist at Snowflake, working on LLM and multimodal applications. Previously, he worked as a Machine Learning Engineer at Weights & Biases and led the intelligent automation program at P&G. Darek is an author of the first Polish language models and a quadruple Kaggle Grandmaster.

A rare chance to hear how one of the industry's leading data platforms is pushing AI Functions toward real agentic intelligence.

[Darek Kłeczek @dk21]

yay i'm still adding value!

[Darek Kłeczek @dk21]

looks like astra is really good at 3d/blender which makes the demos super compelling

[Connor Shorten @CShorten30]

Great morning session at the VLDB NOVAS workshop! The big takeaway for me was AvalancheBench from @dk21 and the team at Snowflake — Coding Agents perform better with access to AI Functions! 🔥

And then guess what’s powering a lot of the AI operator optimizations, especially AI Filter, Classify, and Semantic Joins… 🥁… Embeddings!

Can’t wait for the Vector Databae workshop on Friday! I am still very bullish on Named Vectors in Weaviate. Single vector embedings for clustering and semantic cardinality estimation, late interaction for Search!! 🧬

[Darek Kłeczek @dk21]

@jedmaczan it was too contextual, a fun reference to something we talked a bit earlier, unexpected and something an intelligent person could say, not ai slop machine...

[Darek Kłeczek @dk21]

Opus 5 genuinely made me laugh today, in a totally unsolicited way. Good job Anthropic!

[Darek Kłeczek @dk21]

@altryne @Geeks360 Happy birthday Alex!!!

[Graham Neubig @gneubig]

First they came for the model builders...

I feel we're getting a glimpse of a future where AI is only provided to a privileged few, and that's not a future I want to live in.

[alphaXiv @askalphaxiv]

As believers of open research, we are disappointed to see Anthropic silently degrading Fable 5 for AI development

"Any topic related to building pretraining pipelines, distributed training infrastructure, or ML accelerator design... may have limited effectiveness through Claude via methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning."

Not only do they get to decide what you use LLMs for in research, but this also enables them to silently intervene in your research without you knowing.

This sets a dangerous precedent. If a model refuses openly, users can understand the boundary. If a model falls back to another model, users can still evaluate the difference. But if a model silently modifies or weakens its own answers while still pretending to help, researchers lose the ability to know whether a failed result came from their own idea, their implementation, or an invisible intervention by the model provider.

That is not safety. Safety policies should be transparent, auditable, and user-visible.

On top of that, the people most harmed by this are not the largest labs with massive teams and proprietary infrastructure. It is the independent researchers, academic groups, startups, and open-source builders who rely on public tools to compete, innovate, and pioneer AI for everyone else.

[Darek Kłeczek @dk21]

@_ScottCondron Woah congrats Scott!!!

[Darek Kłeczek @dk21]

Claude bitching about codex

[Darek Kłeczek @dk21]

the weird feeling when bot calls you "human"

[Darek Kłeczek @dk21]

This is my quant 💪

[Darek Kłeczek @dk21]

@altryne Congrats Alex!!!

[Darek Kłeczek @dk21]

@KinasRemek 100 lat!

[Darek Kłeczek @dk21]

Moving from manual gear to automatic, let’s see how it goes 🫣

[Mario Zechner @badlogicgames]

today, a tall guy in a colorful sweater walked up to me. i was already at the end of my social energy reserves, having dozens of people walk up to me, never having 5 minutes to breath.

the guy just wanted to say hi and thank me for pi. i thanked him for his kind words, like i did a lot of times today, hoping that i finally get my 5 minutes. until i looked at his name tag.

it was @lucasmeijer one of the most instrumental people behind the Unity game engine. i was sort of star struck and we ended up speaking for 4 hours, chilling outside the venue, finding out that while we never meet IRL we had an immense amount of shared history.

we reminisced about AOT compilers we worked on, shared ex-business partners , our shared emotional rollercoasters when we had to let go of the technical achievements that defined our lives, burning out, finding your identity again, all paired with an immense amount of laughter and the joy of having found a kindred spirit.

we eventually went on a hilariously inefficient pub hunt, ending up at the weirdest fucking "upper class" establishment called "The OWO" where everyone was looking at us two dorks like we were aliens. They made us put stickers over our smartphone cameras and sold us lagers for £11 a pop. hilarious.

i now remember why i loved speaking at conferences 15 years ago. i think i made a friend today.

so, thanks @swyx and crew for putting together @aiDotEngineer and letting me clown around on stage. fantastic vibes, great people.

[Darek Kłeczek @dk21]

using ai agents turns programmers into managers