Head of product for AI at Snowflake. Previously Google Now cofounder, Google Assistant product director.
Palo Alto, CA
Joined October 2007
- Tweets70
- Following352
- Followers889
- Likes72
Baris Gultekin retweeted
Yay, our team has just published a new paper, “Shift Parallelism: Low-Latency, High-Throughput LLM Inference for Dynamic Workloads"
arxiv.org/abs/2509.16495
Shift Parallelism is a new inference parallelism strategy that can dynamically switch between Tensor Parallelism and Sequence Parallelism, delivering:
- Up to 1.5x lower latency in interactive workloads
- Up to 50% higher throughput under heavy traffic
The tech shows a robust performance across dynamic, real-world traffic pattern
The creators are: Mert Hidayetoglu, Aurick Zhou, Michael Wyatt, Jeff Rasley, Yuxiong He, and Samyam Rajbhandari @ Snowflake AI Research
This work extends Arctic Inference to further optimize LLM inference under dynamic, real-world traffic patterns. The working code is here: github.com/snowflakedb/Arcti…
Cortex Agents are out, with @AnthropicAI as a key LLM powering it.
➡️ Orchestrate across structured and unstructured data sources
➡️ State of the art retrieval: Cortex Analyst with agentic text-to-SQL for high-quality structured data retrieval. Cortex Search with state of the art search quality.
➡️ All running next to data, within the unified governance and security of @SnowflakeDB
snowflake.com/en/blog/ai-dat…
Excited to share @Snowflake’s One Million Minds + One Platform initiative -- a $20 million commitment to upskill 1M+ people in AI and data by 2029.
Hear more from our @RamaswmySridhar on how we're empowering the next wave of data-driven AI innovators.
🌎 Introducing @SnowflakeDB's One Million Minds + One Platform - where we’re empowering one million minds with AI and data skills 🧠
I couldn’t be prouder to support the next generation of builders, AI users, developers, and data experts with this initiative. 🚀
snowflake.com/en/news/press-…
Snowflake, with Hao AI Lab, placed no. 1 on the Spider 2.0 leaderboard, a new benchmark that offers a more realistic test of LLMs’ performance on complex enterprise tasks. Exciting possibilities ahead.
Wow, congratulations 🎉! A team achieved a performance of 24.68% on Spider 2.0-Snow. Are there any better methods out there? 🧐
spider2-sql.github.io
Snowflake customers get immense value from the open source models we host in Cortex AI. For example, open source models like @AIatMeta’s Llama have proven to be highly capable and suited for enterprise-grade AI work. I see it first-hand every day.
Applied AI: Why Businesses Are Skipping Open-Source Models
Read more from @jon_victor_ 👇
Exciting win for open source embeddings - Arctic Embed 2.0 is out, state-of-the-art multilingual embeddings for retrieval tasks, with high inference efficiency.
Available as part of Snowflake Cortex or on HuggingFace under an Apache-2 license.
We’re thrilled to announce the release of Arctic Embed 2.0, the next iteration of our frontier embedding models.
Arctic Embed 2.0 builds on the robust foundation of our previous releases, adding multilingual support without sacrificing English performance or scalability.
“Multilingual embedding models are crucial for enabling people worldwide — not just English speakers — to become AI builders,” said CEO of Hugging Face @ClementDelangue. “By releasing these state-of-the-art models as open source on Hugging Face, Snowflake is making a tremendous contribution to AI and the world.”
Learn more about Arctic Embed 2.0, available on @huggingface: okt.to/1xyqsb
Baris Gultekin retweeted
🚀 I am thrilled to introduce @SnowflakeDB 's Arctic Embed 2.0 embedding models! 2.0 offers high-quality multilingual performance with all the greatness of our prior embedding models (MRL, Apache-2 license, great English retrieval, inference efficiency) snowflake.com/engineering-bl…🌍
Baris Gultekin retweeted
Get up to speed on our latest AI innovations 🏎️
Our Head of AI @barisg joins @jacobeffron on the Unsupervised Learning podcast for a deep dive into our AI ecosystem 🎧
Baris Gultekin retweeted
Our Head of AI @barisg joined the @NVIDIAAI podcast to discuss how we are using AI to enable customers to access and manage data at scale.
By separating the storage of data from compute, Snowflake allows organizations across the world to connect via cloud technology and work on a unified platform — eliminating data silos and streamlining collaborative workflows.
Listen now 🎧: blogs.nvidia.com/blog/podcas…
#Snowflake Cortex Analyst is now available, an agentic AI system for self serve analytics, achieving a remarkable 90%+ SQL accuracy, almost 2x better in text-to-SQL accuracy than GPT4o.
venturebeat.com/data-infrast…
It was great to chat with the awesome @ttunguz about enterprise AI. Thank you for hosting me Tomasz!
Our Head of AI Baris Gultekin (@barisg) sat down on the Office Hours with Tomasz Tunguz (@ttunguz) podcast to dig into what Snowflake’s been up to in the AI space — including a debrief of our latest Llama announcement, how our world-class AI Research Team is pushing the boundaries of innovation, Snowflake’s shift to developing in the open, and more.
Baris Gultekin retweeted
Our Head of AI Baris Gultekin (@barisg) sat down on the Office Hours with Tomasz Tunguz (@ttunguz) podcast to dig into what Snowflake’s been up to in the AI space — including a debrief of our latest Llama announcement, how our world-class AI Research Team is pushing the boundaries of innovation, Snowflake’s shift to developing in the open, and more.
#Snowflake's genAI platform Cortex is now generally available, with a large set of models from Snowflake, Facebook, Mistral and Reka.
Industry-leading LLMs, Generative AI in SQL Functions, and Improved retrieval is now generally available with #SnowflakeCortex 🎊
Baris Gultekin retweeted
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 ...
Another #SnowflakeArctic update - our document focused model beats #GPT4 in document question answering with an orders of magnitude smaller (so cheaper) model!
Excited to release #SnowflakeArctic, a state-of-the-art open source foundation model for enterprise tasks like SQL generation, coding, and instruction following.
snowflake.com/blog/arctic-op…
And #SnowflakeArctic is truly open, with an Apache 2.0 license, and all of our #data recipes and cook books.
#Snowflake is open-sourcing world's best practical embedding model, beating the performance #OpenAI and Cohere text embedding APIs 🚀
These state-of-the-art models, built with Neeva's expertise in search, are now available to download on #HuggingFace.
snowflake.com/blog/introduci…
Reka core model is out with impressive metrics! Excited to partner with @DaniYogatama and the @RekaAILabs team to bring Reka core to @SnowflakeDB #Cortex.
Snowflake Copilot, world’s most capable SQL assistant is now available for everyone. Thank you @MistralAI for the close collaboration to achieve state of the art quality!
Snowflake Copilot, a breakthrough AI-powered SQL assistant powered by @MistralAI large is now in public preview @SnowflakeDB!