Michał Gdak

One engineer’s chief takeaway after his first year at Snowflake? It’s the sweet spot between building big and moving fast. From major projects to massive launches, get Principal Engineer Staszek Pasko's 360° view on the first 365 days at Snowflake here: bit.ly/4pYmIcf

Want to know how Snowflake made the artic-embed models better than Open AI's text-embed-3 or Cohere's embed 3? Check out our tech report on Arxiv arxiv.org/abs/2405.05374. 🧵of TLDR findings below.

3rd week in a row, 3rd LLM from @SnowflakeDB ... Arctic-TILT is a 800M model that has GPT-4 quality performance on information extraction tasks, as measured by the DocVQA benchmark. And it fits in an A10!

World-class search needs world-class metrics 🚀 And great metrics need to be constantly evolved to avoid overfitting 🤟 As we build out a great search experience at @SnowflakeDB AI research, we are excited to join forces with @lintool and the University of Waterloo 🙌 Our focus will be on BEIR v2, a rethink of MTEB/BEIR, and TREC-RAG ...

1/4 Have you wondered how to optimize sys-perf for training Arctic-like models (MoE arch)? Let’s dive in! Our first technique: custom fused kernels. By crafting these kernels, we streamline irregular and sparse operators, boosting efficiency. #SnowflakeArctic #SystemOptimization

A lot of the insider knowledge on how to build an LLM has gone underground in the last 24 months. We are going to build #SnowflakeArctic in the open Model arch ablations, training and inference system performance, dataset and data composition ablations, post-training fun, big run stability, decontamination tricks, metric subtleties. To start with the first two blogs in our cookbook, keep reading the 🧵below ...

<marketing_message> Wonder how to filter web data for high quality? Wonder how you compose the code part of your training data? Read on for what @nathanwiegand and friends did as part of building #SnowflakeArctic ... </marketing_message>

A lot of the insider knowledge on how to build an LLM has gone underground in the last 24 months. We promised to build #SnowflakeArctic in the open, and here we are, with the third edition of our cookbook series, this time on data ... Data ablations are the lifeblood of any LLM training run, whether processing, filtering, deduping, data composition, or schedule. To read about our data approach to Arctic, see the blog below

Introducing Snowflake Arctic. An efficiently intelligent and truly open LLM built by Snowflake.

Beating GPT-4 is not an easy task but with a great team of researchers at @SnowflakeDB we managed to get SOTA on a real text-to-SQL task 💪 Benchmarks available on the market for text-to-SQL are way too trivial for real business usage.

Snowflake announced our intent to acquire Applica.ai to enable document understanding, which is part of our strategy to mobilize unstructured data for advanced analytics and machine learning. See the blog: snowflake.com/blog/snowflake…

ArXiv Highlights: "STable: Table Generation Framework for Encoder-Decoder Models" by @MichaPietruszka, Michał Turski, @lukaszborchmann, et. al Frames table structure recognition as an end-to-end seq2seq task with constrained output. arxiv.org/abs/2206.04045

Any admin interested to work with the biggest farm of GPUs in Poland from nVidia incl. Tesla A100 with 80GB RAM? If so please apply or text me #work #engineering #jobs lnkd.in/eiY98AM

Proud to be part of the company 🏢 named as one of the most promising AI 🤖 based startups in Poland by Rzeczpospolita lnkd.in/eMbyGYv lnkd.in/e8ajWbv

A must read book 📖 for anyone working with technology 👨‍💻 which means almost everyone 😎 Great plot and action. As good as a thriller story 😉 #worthreading #readinglist #goodbook lnkd.in/eatXBuu

We are still hiring. If anyone is thinking of making next career move get in touch with me 😎 lnkd.in/ekRg5jB

I'm really proud to be part of the team 😎 Looks like really big players recognized our technology and efforts to change process automation business by levaraging NLP #nlp #gartner #processautomation lnkd.in/gGm4xY9

I'm happy to be part of the group which gathers & analizes COVID data on a daily basis. Thanks @antoniogulli for the initial idea which has spreaded so fast #corona19 #isolation #zostanwdomu #lockdown #stayhome #coronavirus lnkd.in/gZZ6H9N

I'm excited to join Applica and embark on a journey of #AI / #NLP 🤖 democratization as Head of Product Development. Stay tuned to see our product shortly 😎 #artificialintelligence #development #learning #newjob #newjourney lnkd.in/gqzERHN

For those who couldn't participate in last Kaggle Days meetup 🤖 in Abu Dhabi U can watch 📽️ my speach 🎙️ on YT Thank @ajaypundhir for the opportunity #kaggle #development #learning #knowledgesharing lnkd.in/gUP_JJd lnkd.in/ghgcx9n