AI Summit West: Deep Learning & Advanced ML schedule

• Discuss the ethical considerations surrounding AI development and deployment.  

• Learn about best practices for building fair, transparent, and accountable AI systems.  

• Explore the role of governance and regulation in promoting ethical AI. 

Moderator: Lisa M. Lum, Director, Board of Directors, Cal Alumni Association | UC Berkeley 

- Abhai Pratap Singh, Senior Product Manager-Technical, Amazon Alexa 

- Sayan Maity, Principal ML Engineer, Disney Streaming  

- Jitender Jain,  Former Principal Software Engineer, Capital One

In the AI race, data can be your most valuable asset, but without robust data-to-AI governance, your innovation is built on shaky ground. This talk exposes the hidden risks of neglecting governance, from biased models to legal pitfalls, threatening your AI initiatives. Learn how a strong data-centric governance framework not only mitigates these risks but fuels responsible AI, enabling faster development, increased trust, and more impactful results.

To follow ... 

• Key insights and strategies from working with enterprise clients to build effective retrieval-augmented generation (RAG) systems. 

• Practical tips for designing “Minds” for optimal retrieval performance. 

• What worked, what didn’t, and how to tackle common challenges. 

To follow ... 

• Explore how financial services and enterprises can leverage AI as a strategic asset to gain a competitive edge. 

• Learn about tailored strategies for designing and implementing successful AI initiatives in financial services and broader business contexts. 

• Discover real-world examples of companies leveraging AI for strategic advantage. 

As Machine Learning continues to forge its way into diverse industries and applications, optimizing computational resources, particularly memory, has become a critical aspect of effective model deployment. This session, "Memory Optimizations for Machine Learning," aims to offer an exhaustive look into the specific memory requirements in Machine Learning tasks, including Large Language Models (LLMs), and the cutting-edge strategies to minimize memory consumption efficiently. 
 
We'll begin by demystifying the memory footprint of typical Machine Learning data structures and algorithms, elucidating the nuances of memory allocation and deallocation during model training phases. The talk will then focus on memory-saving techniques such as data quantization, model pruning, and efficient mini-batch selection. These techniques offer the advantage of conserving memory resources without significant degradation in model performance. 

A special emphasis will be placed on the memory footprint of LLMs during inferencing. LLMs, known for their immense size and complexity, pose unique challenges in terms of memory consumption during deployment. We will explore the factors contributing to the memory footprint of LLMs, such as model architecture, input sequence length, and vocabulary size. Additionally, we will discuss practical strategies to optimize memory usage during LLM inferencing, including techniques like model distillation, dynamic memory allocation, and efficient caching mechanisms. 

By the end of this session, attendees will have a comprehensive understanding of memory optimization techniques for Machine Learning, with a particular focus on the challenges and solutions related to LLM inferencing. 

LLMs have started providing great value to enterprise use cases. Deploying smaller LLMs in production is attractive since they are less resource hungry. Smaller LLMs, though efficient, often struggle to match the performance of their bigger counterparts. In this talk, we will discuss various techniques like knowledge distillation, post-training alignment like RLHF and DPO, etc. that can help bridge this gap. The speakers will discuss success stories from open source and enterprise applications. 

Speakers: 

- Sandeep Jha, Principal Staff Technical Program Manager, LinkedIn 

- Aman Gupta, Sr. Staff Engineer - Applied AI Research, LinkedIn 

• Understand the best practices for scaling, deploying, and managing Gen AI agents in enterprise environments. 

• Learn how to address challenges related to security, compliance, and performance optimization in production. 

• Gain actionable strategies for leveraging Gen AI to drive innovation, improve operational efficiency, and create business value. 

• Acquire insights on operationalizing, optimizing, and efficiently scaling deep learning model training. 

• Learn from case studies, on managing ML platforms with web-scale data. 

• Understand how modern ML computing frameworks like Ray and PyTorch can be utilized to create impactful ML products. 

• Dive into deep technical discussions regarding ML/AI and infrastructure 

• AI-Driven 3D Content Creation: Showcase how machine learning algorithms automate and enhance 3D modeling (Discuss about NeRF) 

• AI-Driven Occlusion for XR Applications: Highlight how AI models, such as those derived from ARKit and ARCore, enable accurate object and people occlusion, which is critical for creating immersive and realistic XR experiences. 

• Deep Learning for Order-Independent Transparency (OIT): Explore how deep learning techniques can enhance traditional transparency algorithms like k-buffer, improving performance and visual quality in real-time applications where transparency is essential. 

• Achieving the Unimaginable: See how AI enables the creation of thousands of real college football players in College Football 25, enhancing player experiences at scale. 

• Breaking Barriers with GPTs: Gain actionable strategies to expand into new markets and engage diverse audiences, including neurodiverse customers, using advanced generative AI tools. 

• Navigating AI Risks: Understand the pitfalls of bias in large AI models and learn best practices to mitigate its impact for more equitable and ethical outcomes. 
  

 Speakers: 

- Faith McGee, Sr. AI/Analytics Research Manager, Electronic Arts 

- Natashia Tjandra, Research Director, Electronic Arts

Prithvi Srinivasan and Shih-Ting Lin will explore how Instacart leverages Large Language Models (LLMs) to extract and manage information from raw product data within Instacart's extensive catalog, which includes millions of products from stores across the US and Canada. Their presentation will cover the process of extracting essential product attributes, the effective utilization of these attributes, and the tooling developed for backend engineers to streamline these operations. 

Speakers: 

- Prithvi Srinivasan, Machine Learning Lead, Instacart  

- Shih-Ting Lin, Senior Machine Learning Engineer, Instacart 

Personalization has become the cornerstone of innovation across industries, driven by AI’s ability to understand, anticipate, and respond to individual user needs at scale. From financial services creating tailored investment plans to healthcare providing personalized treatment recommendations, AI enables hyper-relevant experiences that delight users and deliver measurable results. This talk explores how AI-powered personalization is transforming the way businesses deliver value and build stronger connections with users across diverse sectors.