The Ultimate Recap of Snowflake’s Dev Day
Didn’t make it to Dev Day? Fear not, we’ve recapped everything that took place. It will almost feel like you were there, and hopefully we’ll catch you at the next one!
TL;DR on June 6th 2024, Snowflake hosted its first ever Dev Day, a free, all-day event on the final day of Snowflake Summit. Dev Day brought together enthusiasts in the AI and apps space to learn from visionaries, level-up their skillset, and meet peers and mentors alike.
Here’s what happened…
The Startup Challenge Finale
Dev Day kicked off with the Startup Challenge Finale, where three remarkable finalists took the stage, vying for up to $1 million in total investments. Each startup showcased their cutting edge solutions, and an all-star panel of judges including Benoit Dageville, Brad Gerstner, Denise Persson, and Lynn Martin selected the winner. Kudos to SignalFlare.ai for coming in first!
Following the Challenge, leading VCs, including Pauline Yang from Altimeter and Philip Clark from Thrive Capital, took the stage to share insights into the latest trends in data and AI, and what they look for when funding new startups. The panelists discussed how AI startups can build technical advantage into their foundational architecture by focusing on two key elements: data gravity and distribution.
Dev Day Luminaries
On the main stage the focus shifted from startups to luminaries, opening with a guest appearance from Snowflake CEO, Sridhar Ramaswamy, to share Snowflake’s commitment to advancing AI technology.
“We are opening up; we are becoming an application platform with a healthy dose of Open Source and community-led development.”
Sridhar then invited Dr. Andrew Ng, founder and CEO of Landing AI, to join him on stage for a discussion about his journey in the AI space, and where the industry is headed, with a focus on agentic AI.
“AI agents significantly expand the set of what can be done with AI. If I had to pick one thing for any builder to keep an eye on, it would be AI agents.”
The Luminary Talks continued with Weights & Biases CEO and founder, Lukas Biewald, emphasizing the current trend of democratizing AI, where AI capabilities can be easily integrated by any developer. He discussed the significant challenge of transitioning AI from impressive demos to reliable production systems, highlighting the lack of robust developer tools for AI.
“I think executives at every company are frustrated about this right now, I think every developer is frustrated by this…it’s incredibly easy to make an awesome AI demo and it’s so much harder than you would think to get that in production.”
Richard Socher, CEO of You.com and one of TIME Magazine’s most influential people in AI, took the stage next to discuss the evolution and trust issues of large language models (LLMs). He recounted his journey from early multitask models in 2018 to advanced models like ChatGPT, highlighting the challenges in deploying these models in production environments. Socher emphasized the need for enhancing LLMs with real-time internet access, using the best model for specific tasks, and implementing accurate citation and verification methods.
“If it takes you longer to verify the artifact that gen AI put out than it is to create that artifact, then gen AI is probably not a good use case for your organization.”
In her talk on paving the way for voice mode, Christine McLeavey, Head of Audio Research at OpenAI, discussed the breakthroughs and challenges in developing voice mode for ChatGPT, including the recently launchyed GPT-4o. McLeavey previewed upcoming advancements that promise more natural and responsive voice interactions in ChatGPT, underscoring the importance of real-time processing and the model’s ability to understand nuances in human speech.
Kari Briski, Vice President of AI Models, Software and Services at NVIDIA, led a session on the complexities of deploying generative AI, with focus on the rapid evolution and rising computational demands, especially since the introduction of the transformer architecture.
“The introduction of the transformer architecture really marked this paradigm shift by harnessing unlabeled data sets.”
Briski also discussed NVIDIA’s efforts to simplify AI model deployment through their AI Foundry and NIMs (NVIDIA Inference Microservices), which package optimized models for easier and more efficient production use.
For the final Luminary Talk of the day, Yusuf Ozuysal, Principal Software Engineer at Snowflake, and Doris Lee, Product Manager at Snowflake, discussed the benefits and considerations of fine-tuning open source LLMs versus leveraging existing model APIs. They covered key topics such as addressing model bias, ensuring code security and observability, and building scalable, open source generative AI applications. Watch the session playback here.
What Else Happened at Dev Day?
While the Dev Day stage was the main attraction, attendees immersed themselves in an array of other activities designed to help them make new connections and level-up their skills.
Gen AI Bootcamp
Created and hosted by Dash Desai and Julian Forero, the Gen AI hands-on lab instructed participants how to combine data with LLMs using Snowflake Cortex and serverless LLMs. Attendees learned to build and deploy scalable, secure apps, earning a custom bootcamp badge upon completion.
If you’re interested in upskilling, you can take a related course on Coursera and earn your own badge!
connect() Meetups
Networking opportunities were abundant, with hourly meetups covering startups, open source, AI, data engineering, and women in data. These sessions facilitated meaningful connections between attendees, and sparked ideas for collaboration. Shoutout to the amazing local communities who showed up in full force, including Data Engineer Things, WiMLDS, PyBay, and our very own Streamlit community!
AWS DeepRacer Track
Adding a fun twist to the day, the AWS DeepRacer Track allowed attendees to race AI-powered cars, combining entertainment with learning about machine learning.
Demo Dare Finale
The fast-paced “5-Minute Demo Dare” finale brought sessions to a close, during which top presenters from our streaming series competed live in front of a packed theater. Hosted by Felipe Hoffa, these finalists demonstrated exceptional skill and creativity, presenting their products and Snowflake’s latest features in just five minutes. Judged by the live audience on engagement, ambition, and usefulness, the event was a showcase of technical brilliance.
[Code Block] Party
Dev Day wrapped up with the [Code Block] Party, where attendees had a chance to network, hang out, and say farewell to newfound friends.
Thank you to everyone who joined us for Dev Day, and a special thanks to our partners who made Dev Day possible:
Astronomer, AWS, Data Science Dojo, General Assembly, HackerRank, NumFocus, Pinecone, Posit, Predibase, PyTorch, Tobiko, Tonic, Voltron Data, and Weights & Biases
What next?
If you missed Dev Day, you can join us virtually for Snowflake BUILD 2024 to connect with developers, data scientists, and engineers and explore Snowflake’s latest innovations. Be sure to also check out Snowflake For Developers, the Solutions Center to discover step-by-step technical QuickStart guides, code snippets, and demo videos, and the Snowflake Community where you can learn, innovate, and problem solve with your peers.
Stay tuned for details on Snowflake’s next Dev Day!
