Fully Connected 2023
San Francisco Conference
Join us for a one-day event focused on the generative models and LLMs transforming the world today!
You’ll hear from the creators of the most exciting LLM tools like LangChain & OpenAI, people building your favorite platforms like AWS, NVIDIA, Lightning, HuggingFace, Kaggle, and companies having a positive impact on the world through machine learning.
At weights & Biases FullyConnecteD, Spring 2023
Bringing Together Builders of Large Language Models
At weights & Biases Fully ConnecteD ON June 7
Join us for a one-day event focused on the generative models and LLMs transforming the world today! You’ll hear from the creators of the most exciting ML tools like LangChain & OpenAI, people building your favorite platforms like AWS, NVIDIA, Lightning, HuggingFace, Kaggle, and companies having a positive impact on the world through machine learning.
We’ll also be launching an exciting new product at Fully Connected that will dramatically change how you train and deploy your models. We will be demoing it live and we can’t wait to hear what you think.
building 12 at Pier 70 in San Francisco
Welcome and Opening Remarks
Lukas Biewald, CEO and Co-Founder at Weights & Biases
We’ll kick off Fully Connected 2023 with a keynote from W&B’s CEO and Co-founder Lukas Biewald. He’ll set the stage for the day and introduce a new product we think you’re going to love.
W&B Product Announcement
Shawn Lewis, CTO and Co-Founder at Weights & Biases
W&B for Model CI/CD
Carey Phelps, Founding PM at Weights & Biases
Join Carey Phelps as she showcases how our newest products and features fit into your current ML workflows. This will be a true end-to-end demo condensed into a brief 15 minutes.
During our main sessions, you’ll hear from speakers from AWS, Langchain, NVIDIA, You.com, Nomic AI, and more, with a special keynote presentation by Tim Urban. Check the “Main Stage” tab for the full list of presentations.
The second half of Fully Connected will be short, rapid-fire breakouts from some of the most innovative voices and companies in machine learning. We have speakers from Lightning AI, Run AI, Mosaic ML, Humane, AWS, Snowflake, Anyscale, Domino, fast.ai, and a whole lot more. Check the “Coatue Breakout Room” and “Felicis Breakout Room” tabs for the full list of presentations.
Afterparty and Networking
Our last few hours will be completely devoted to networking, mingling, and getting to know likeminded ML practitioners. Drinks on us.
Memory in LLM Applications
Harrison Chase, Co-Founder & CEO at Langchain
This talk will focus on how “memory” in the context of LLM applications. Memory is most often discussed in the context of chatbots so it will start with an overview of conversational memory. This can vary from simple (buffer of recent messages) to complex (extracting and dynamically generating a knowledge graph). It will then cover recent advances in memory, including the “Generative Agents” paper. It will finish will an overview of where the space is now and where it could go in the future (it is still early days!)
NLP is the future of AI
Richard Socher, Founder & CEO at You.com
This session delves into the immense potential of natural language processing and generative AI, in revolutionizing summarization, human productivity and online search. Led by Richard Socher, a top NLP research scientist and founder and CEO of You.com, the session showcases his transition from academic research to large-scale AI in industry. Richard will explore AI in search and demonstrate its most impactful use cases.
Building trustworthy, safe, and secure LLM conversational systems
Jonathan Cohen, VP of Applied Research at NVIDIA
Large language models (LLMs) are incredibly powerful and capable of answering complex questions, performing feats of creative writing, developing, debugging source code, and so much more. Yet, building these LLM applications in a safe and secure manner is challenging. Because safety in generative AI is an industry-wide concern, NVIDIA developed NeMo Guardrails as an open-source toolkit to help developers guide generative AI applications to create impressive text responses that stay on track and on topic.
Custom Chatbots using LLMs, Streamlit, LangChain, and Snowflake - Why and How
Adrien Treuille, Head of Streamlit at Snowflake
Chatbots are not just a passing fad. Their ability to express successive task refinement with a Large Language Model (LLM) makes chatbots a fundamental UI paradigm of the LLM era. In this talk, we’ll preview a new chat UI in Streamlit, explain some of the design decisions that went into this design. Finally, we will build our own custom chatbot, live on stage, using Streamlit, LangChain, and Snowflake.
The Future of Generative AI
Ari Kalfayan, global head of early-stage Business Development for AI/ML startups & VC at AWS
Generative AI is one of the hottest topics in business and the AWS Gen AI Accelerator has picked the top 21 Gen AI companies in the world to accelerate their growth and technical development.
Join us for a panel with 4 of the top startups in our inaugural cohort as we talk through the Generative AI market as a whole. We’ll discuss the tooling needed to develop Gen AI applications and share helpful insights on how people can leverage these tools and applications.
- Bernardo Aceituno, Co-Founder at Stack AI
- Eiman Ebrahimi, CEO at Protopia AI
- Mihail Eric, Co-CEO at Storia AI
- Max Mergenthaler, CEO and Co-Founder at Nixtla
Leadership needs us to do gen AI, what do I do?
Chip Huyen, Founder & Ceo at Claypot AI
Everywhere we go, we see companies trying to figure out their generative AI strategy, but few seem to know what exactly to do. In this short talk, we’ll discuss key considerations when evaluating LLMs.
Why ‘no-code’ will power the long tail of AI applications
Howie Liu, Co-Founder & CEO at Airtable
Generative AI is capable of advanced reasoning and creativity that can be immediately applied to a broad range of knowledge work. Yet the bottleneck to broader adoption will be enabling the creation and experimentation of AI-enabled workflows on a large scale. The no-code approach to app-building is very well-suited to enabling the long tail AI applications. In this talk, we discuss the opportunity, and design patterns, for no-code LLM apps.
Large Models, AI, Human History and the Future
Tim Urban, Writer, Illustrator, and Co-Founder of Wait But Why
Back in 2014, long before Artificial Intelligence became the subject that everyone is talking about, Tim Urban began some serious research. It hit him pretty quickly that what’s happening in the world of AI is not just an important topic, but, by far THE most important topic for our future. After three years of diving deeply into the subject and all its future implications, Tim published “The AI Revolution and the Road to Superintelligence.” Vox dubbed this Tim’s “epic series on artificial intelligence,” taking readers on a deep exploration of what AI is, how it works, and why it will dramatically change our lives.
As AI has expanded into our world, so has Tim’s knowledge and insights. With today’s conversations on AI taking on multiple dimensions, Tim Urban remains the ultimate (and frequently entertaining) explainer of the most transformational shift in the way we live, work, and evolve as a society.
Why ML Engineers are still useful when using LLM APIs
Hamel Husain, Independent Consultant / Founder
It’s tempting to feel that ML Engineers are obsolete in the case where you are just using OpenAI’s APIs off the shelf. Even though you don’t need to train or serve a model, evaluation of LLMs are still critical. Thinking rigorously about evaluation metrics and building offline evaluation harnesses are key to enabling rapid development velocity and making progress on LLMs. In this lighting talk, Hamel will share his experiences on this subject.
Using modern NLP to build a knowledge graph
Anthony Goldbloom, Co-Founder & CEO at Sumble; Co-founder & Former CEO at Kaggle
Best Practices of W&B and Amazon Sagemaker
James Yi, AI/ML Tech Lead at AWS + Matt Hoffman, ML Engineer at Weights & Biases
Learn best practices for setting up Amazon Sagemaker with Weights & Biases MLOps platform. Quick demo will show how to setup experiment jobs, model monitoring, check model performance, identify best models in dashboard, and launching training jobs in Sagemaker.
Decrease PyTorch Model Load Times with CoreWeave’s Tensorizer
Rahul Talari, Sr. Machine Learning Platform Engineer at CoreWeave + Navarre Pratt, Machine Learning Engineer at CoreWeave
Large language models and machine learning models with billions of parameters can take a long time to load, severely impacting a company’s ability to quickly scale its inference stack to meet rising demand. To help reduce latency, improve throughput, and reduce resource usage, CoreWeave employs a range of open-source tools, including CoreWeave’s Tensorizer. Join Rahul Talari and Navarre Pratt, machine learning engineers at CoreWeave, as they walk through how CoreWeave’s Tensorizor makes it possible to load models extremely quickly from HTTP/HTTPS and S3 endpoints for better end-to-end latency and reduced resource utilization, and lower costs.
Responsible Generative AI
Josh Mineroff, Head of Technical Alliances at Domino Data Lab
Model explainability is one thing – increasingly, that’s not the hard part. It’s easy to get lost in the technical complexity of LLMs, as well as the abstract concepts surrounding responsible AI. But like any new technology, people and processes matter more than algorithms – and full tracking and reproducibility across the model lifecycle is foundational. Join this breakout session to learn how enterprises are adapting foundation models to meet their business and governance objectives – from tuning to deployment to monitoring.
Customer Trends for Azure OpenAI Engagements
Michael Lanzetta, Lead Technical Architect for Data and AI at Microsoft
As our enterprise partners go from ideation to production, we’re gaining a better understanding of what is resonating with them, their open concerns, and the blind spots we can help illuminate. In this short talk I’ll share what we’re learning, and which insights I believe are evergreen and which are rapidly evolving.
State of GPU Market for Deep Learning
Mitesh Agrawal, Head of Cloud at Lambda
- the current availability of GPUs
- challenges in the supply chain, and
- the implications of large-scale deployments.
LLMs ❤️ Enterprise Data - Perils and Opportunities
Venkat Krishnamurthy, Head of ML Platform at Snowflake
LLMs and Generative AI are being talked about as equivalent to the Industrial Revolution, in terms of potential impact to society at large. The enterprise world is no exception. We’ll cover the promise of LLMs when applied to enterprise data, but also uncover some obvious, and some not so obvious perils along the way.
Responsible AI for Enterprise Adoption of Generative AI
Ali Arsanjani, Director of Cloud Partner Engineering at Google
- Generative AI offers enterprises opportunities for creativity, innovation, and automation, but responsible AI practices are crucial for ethical and unbiased outcomes.
- Understanding generative models and data requirements, including diversity and privacy concerns, is essential for successful enterprise adoption of generative AI.
- A responsible AI framework for enterprise adoption involves ethical considerations, transparency and explainability, robustness and security measures, and ensuring legal and regulatory compliance.
- Mitigating social impact and bias in generative AI outputs, along with addressing the impact on human labor and creativity, are crucial considerations for responsible enterprise adoption.
- Addressing the adaptation of foundation models through AI safety, toxicity mitigation, misuse prevention, and robustness is paramount to ensure ethical and reliable outcomes in enterprise applications.
The Science Behind MPT-7B, Our Open Source LLM
Hanlin Tang, Co-Founder & CTO at MosaicML
Orchestrating and Tracking the Full AI Lifecycle: Training LLMs and Serving Inference
Robert Magno, Solutions Architect at Run:AI
What is GPT-4 good for?
Ted Sanders, Member of Technical Staff at OpenAI
As a research engineer at OpenAI, I advise companies, from startups to Fortune 100, trying to build GPT into their products.
This talk will share my perspective on:
- GPT’s #1 superpower
- GPT’s #1 weakness
- The #1 misconception about GPT
- The #1 mistake teams make when building products atop GPT
- The markets and applications I’m most excited by
Accelerating AI Product Development with Lightning AI
William Falcon, Founder & CEO at Lightning.AI
Join us for an engaging community talk on Lightning AI, designed by the creators of PyTorch Lightning. Lightning AI is the platform to build foundation models, on your data, your cloud. Lightning expedites AI product development by providing a comprehensive suite of features to build, train, deploy, and ship AI products seamlessly.
During this talk, we will explore how Lightning AI simplifies and streamlines the entire AI development process. By enabling distributed training, hyperparameter tuning and autoscaling at the click of a button, the Lightning platform empowers developers to focus on their models and ideas rather than dealing with the underlying infrastructure.
Through real-world examples and use cases, we will demonstrate how the Lightning Platform significantly reduces the time and effort required to build and deploy AI products. Whether you are a researcher, data scientist, or AI enthusiast, this talk will equip you with the knowledge and tools necessary to accelerate your AI development workflow and maximize productivity.
Join us for an insightful session filled with valuable insights, practical tips, and a glimpse into the future of AI development.
PatentPT: Building an LLM-Powered Solution with Enterprise-Grade Memory Agents
Davit Buniatyan, CEO & Founder at Activeloop
Learn how we created PatentPT, an advanced language model solution that greatly enhances patent search and interaction capabilities. The presentation offers practical insights on fine-tuning and deploying large language models and leveraging enterprise-grade memory agents to autocomplete patents, generate abstracts and claims, and conduct advanced patent search functions using the rich patent corpus. We’ll walk you through how to develop a similar solution using cutting-edge Activeloop’s Deep Lake, the Database for AI, open-source LLM models, Habana Gaudi HPU hardware, and Amazon Sagemaker’s LLM inference APIs.
We will walk you architectural blueprints and all the steps we took to build the solution – from training our LLM model, and finetuning it, as well as creating custom featurizers and the deployment of search APIs.
How LLMs have redefined the way we do AI
Robert Nishihara, Co-Founder & CEO at Anyscale
Large Language Models are getting more attention as a way to address the data privacy, latency, flexibility, and cost challenges of using general purpose LLM services. But deploying LLMs in production means solving for performance, scale, infrastructure/cloud costs, and integration with other tools and frameworks. In this session, you’ll hear about what Anyscale has learned from powering the world’s most popular LLMs including OpenAI, and how Anyscale helps organizations productionize and scale LLMs.
Towards Explainable and Accessible AI
Brandon Duderstadt, Founder & CEO at Nomic AI
Building AI with open source and Hugging Face
Jeff Boudier, Product Director at Hugging Face
In this talk, Jeff Boudier will tell you about the latest and greatest releases in the beautiful world of open source machine learning, and walk you through how you can build AI into your apps using Hugging Face tools.
Please make sure you enter the code in the correct space (you’ll see a section marked “X Promo code”) and click on Apply. If this isn’t working, your code might be expired or incorrect. If you keep experiencing issues, please contact us at email@example.com.
Yes! You can transfer your ticket to another person by sending an email to firstname.lastname@example.org and we will update the ticket contact details for you.
Please check your spam or junk folder to ensure the confirmation email didn’t end up there. If you still haven’t received it, contact our support team at email@example.com with your registration details and we’ll assist you.
Yes, that’s correct. There will be alcohol served after 5. To ensure compliance with legal regulations and promote responsible drinking, the conference organizers have implemented an age restriction policy, allowing only individuals who are 21 years or older to attend and partake in these activities.
Entry and Check-in
Please show the QR that you’ll receive 24 hours before the event for smooth check-in. Additional requirements or instructions will be communicated prior to the event if necessary.
The in person event is at Pier 70, Building 12, San Francisco, CA. You will find more details in the email going out 24 hour before the event. We recommend using navigation apps or public transportation options for ease of travel.
Parking is available at the event venue but is limited! We will send an email with specifics for parking details.
Yes there will be snacks & drinks provided between 1-5pm. At 5pm there will be more food, drinks, and time to network with the attendees and sponsors.
This event is run entirely in person. We will record the main keynote sessions but the event will not be live streamed.
Attendees will receive a bag of goodies upon registration.
Additionally, you can participate in the W&B Treasure Trek by scanning QR Codes to unlock clues. Use the clues to guess the right answer to win a pair of Apple AirPods or Bose Speakers. Find the first clue in our Discord: http://wandb.me/discord