Snorkel AI Courses

Mastering LLMs with Skill-it! and Zero-Shot Robustification

Explore innovative techniques for optimizing language models: Skill-it! for efficient training and zero-shot robustification for improved performance. Learn data-driven approaches to enhance AI capabilities.

• YouTube
• 20 minutes
• On-Demand
• Free Video

Data Development for Improving GenAI Systems - A Systems-Level Approach

Explore how data development enhances GenAI systems through a case study improving RAG accuracy by 54 points in 3 weeks using scalable tools for embedding and retrieval models.

• YouTube
• 18 minutes
• On-Demand
• Free Video

Modeling Human Preference to Improve LLM Performance

Explore techniques for aligning language models with human preferences, including reward modeling and scalable tools to improve LLM performance with minimal end-user guidance.

• YouTube
• 20 minutes
• On-Demand
• Free Video

Better not Bigger: Distilling LLMs into Specialized Models

Explore techniques for distilling large language models into specialized, deployable versions while maintaining performance and reducing data requirements.

• YouTube
• 17 minutes
• On-Demand
• Free Video

SQL for Efficient Data Organization in Machine Learning

Explore JoinBoost, a Python library transforming tree training algorithms into SQL queries for efficient ML on normalized databases, offering improved performance and scalability.

• YouTube
• 11 minutes
• On-Demand
• Free Video

Maximize ML Model Performance with Data-IQ Framework

Discover Data-IQ: a framework to stratify data into subgroups, enabling systematic auditing of ML models across tabular, image, and text data for improved performance and reliable usage.

• YouTube
• 12 minutes
• On-Demand
• Free Video

The Ethical Implications of Building a Real-Life Skynet

Explore the ethical challenges and limitations of generative AI, its potential for improvement, and its broader implications in the evolving landscape of artificial intelligence.

• YouTube
• 14 minutes
• On-Demand
• Free Video

How to Operationalize Knowledge for Data-Centric AI

Explore data-centric AI's philosophy and practical implications, focusing on operationalizing knowledge to programmatically label unstructured data for modern AI applications.

• YouTube
• 19 minutes
• On-Demand
• Free Video

Dataset Cartography: What Capital One Monitors to Better Manage AI Training Data

Explore dataset cartography techniques used by Capital One to enhance AI training data management and improve model performance.

• YouTube
• 20 minutes
• On-Demand
• Free Video

LIGER: Fusing Model Embeddings and Weak Supervision for Improved NLP and Vision Tasks

Explore LIGER: a novel approach fusing foundation model embeddings with expert-driven labeling functions to enhance NLP and vision tasks.

• YouTube
• 11 minutes
• On-Demand
• Free Video

Knowledge Distillation Demystified: Techniques and Applications

Explore knowledge distillation techniques for optimizing ML models in NLP. Learn extraction and transfer methods, target skill identification, and innovative data-centric approaches for efficient model training.

• YouTube
• 24 minutes
• On-Demand
• Free Video

Model Distillation - From Large Models to Efficient Enterprise Solutions

Explore model distillation techniques for efficient enterprise AI systems. Learn strategies, real-world applications, and best practices to optimize performance and minimize costs in natural language processing.

• YouTube
• 57 minutes
• On-Demand
• Free Video

Curate Training Data via Labeling Functions - 10 to 100x Faster

Accelerate data labeling for AI models using programmatic techniques. Learn to capture SME knowledge, improve label accuracy, and explore real-world use cases in banking, pharma, and healthcare.

• YouTube
• 57 minutes
• On-Demand
• Free Video

Revolutionizing Data Analysis: Combining LLM, RAG, and Domain Knowledge

Leverage NLP and RAG systems to extract insights from structured and unstructured data, generate actionable insights, and empower data-driven analysis without SQL.

• YouTube
• 51 minutes
• On-Demand
• Free Video

Task Me Anything: Revolutionizing Multimodal Model Benchmarking

Explore Task Me Anything platform for efficient multimodal model evaluation. Learn programmatic task generation and active learning techniques to create custom benchmarks for data science projects.

• YouTube
• 41 minutes
• On-Demand
• Free Video