You can finish all of the courses in an hour
Deep Learning's creator, Andrew Ng.An influential person in the domains of deep learning and machine learning, AI is also a co-founder of Coursera. People hold his AI courses in high respect since they are well-structured and provide insights into the most recent advancements in the area.
Ng's classes often incorporate hands-on projects and assignments that let students practice using deep learning models and algorithms in the real world. To take into account the most current advancements in deep learning, these courses are updated on a regular basis.
These are the most recent courses taught by Andrew Ng to help you learn and improve your AI abilities.
Artificial Intelligence Agents in LangGraph
Learn how to combine agentic search with query-focused, consistently formatted responses to augment an agent's knowledge in this brief course. Additionally, you will discover how to use agentic memory to store state for debugging and reasoning and discover how human-in-the-loop input may direct agents at crucial points.
To fully grasp the framework, one may create an agent from scratch and then rebuild it using LangGraph. Ultimately, a sophisticated essay-writing agent that integrates all of the course material will be developed.
Enroll here and find out more about the course.
AutoGen-Based AI Agentic Design Patterns
This course will teach you how to use agentic design patterns—like multi-agent cooperation, nested and sequential conversation, reflection, tool usage, and planning—using AutoGen.
Additionally, you will discover how to construct and integrate specialized agents—such as planners, researchers, programmers, writers, and critics—that collaborate to carry out intricate processes like producing thorough financial reports that would otherwise need a great deal of human labor.
Key agentic design ideas are covered throughout the course along with enjoyable examples. For example, it is possible to create a conversational chess game with two player agents that confirm moves, update the state of the board, and have animated conversations about the game.
Learn more about the course and make an enrollment here.
Overview of AI on Devices
You will study the critical stages for on-device deployment in this course, which include hardware acceleration, neural network graph capture, on-device compilation, and numerical correctness checking. You will also deploy a real-time image segmentation model on a device.
You will also discover how quantization may improve performance on edge devices with limited resources by making the model 4 times quicker and 4 times smaller. By using these methods, models may be installed on a variety of gadgets, such as robots, drones, and smartphones, opening up a wide range of innovative and creative applications.
Learn more about the course by visiting this link.
Crew AI in Multi-AI Agent Systems
This course will teach you how to divide up difficult jobs into smaller assignments for several AI agents, each with a specialized function.
For instance, collaborating on a study report may need authors, quality assurance representatives, and researchers. Just as with managing a team, one may specify their duties, expectations, and interactions.
Furthermore investigate important AI strategies such guardrails, memory, role-playing, tool use, and cross-agent cooperation. Additionally, create multi-agent systems to handle challenging assignments. Creating and seeing these agents work together is both productive and entertaining.
Enroll here and find out more about the course.
Constructing RAG and Multimodal Search
This course will teach you how to integrate multimodality to RAG so that models may utilize a variety of relevant settings to answer questions, as well as how contrastive learning works.
For example, a financial report query may have tables, slides, graphs, and text excerpts. Additionally, one will discover how to create a multi-vector recommender system utilizing Weaviate's open-source vector database and how visual instruction tuning incorporates picture understanding into language models.
Learn more about the course by visiting this link.
Creating Agent-Based RAG Using LlamaIndex
This addresses a significant change in RAG: instead of requiring the developer to provide specific algorithms for information retrieval for the LLM context, one may create a RAG agent that has access to a variety of information-retrieval tools.
In-depth information will be covered on routing, which involves the agent using judgment to route requests to various tools; tool use, which involves building an interface that allows agents to choose the proper tool (function call) and provide the necessary arguments; and multi-step reasoning using tool usage.
Learn more about the course by visiting this link.
In-Depth Quantization
You will learn how to create both symmetric and asymmetric modes of linear quantization from scratch in this course. To preserve performance, it will also quantize at various granularities (per-tensor, per-channel, and per-group).
You will be able to build a quantizer that can compress any open-source deep learning model's dense layers to 8-bit precision. Lastly, by squeezing four 2-bit weights into an 8-bit integer, you will experience quantizing weights into 2 bits.
Learn more about the course by visiting this link.
In Prompt Modeling for Vision Engineering
In this course, you will discover how to build and optimize vision models for segmentation, object identification, editing, and personalized picture production. Prompts might be bounding boxes, text, or coordinates, depending on the model. To further shape the outcome, hyperparameters will also be adjusted.
Working with models such as OWL-ViT, Segment-Anything Model (SAM), and Stable Diffusion will be taught. Additionally, to refine Stable Diffusion by using a small number of photos to get customized outcomes, such pictures of a certain individual.
Register for the course and learn more here.
Getting Mistral Started
Using API calls and Mistral AI's Le Chat website, you will investigate Mistral's commercial models (Mistral 8x7B, Mistral 7B) as well as its open-source models throughout this course.
Use JSON mode to provide outputs that are organized and can be directly integrated into bigger software systems. Additionally, you may utilize function calling to employ tools, such custom Python code that does tabular data queries.
Using RAG, anchor the LLM's answers to outside knowledge sources. Create a chat interface using Mistral that can link to outside papers. One's prompt engineering abilities will be strengthened by taking this course.
Enroll in the course and get more information here.
Preparing Unstructured Information for LLM
It is crucial to extract and normalize material from many forms, such HTML, PowerPoint, and PDF, in order to increase LLM's expertise. This entails adding metadata to the data to enhance its retrieval and reasoning capabilities.
This course will teach you how to prepare data for LLM applications, with an emphasis on different kinds of documents. Learn how to extract and normalize documents into a standard JSON format that is enhanced with information to improve search engine optimization.
The course covers methods for handling PDFs, pictures, and tables using document image analysis, such as layout identification and vision transformers. You will also learn how to create a RAG bot that can read a variety of documents, including Markdown, PowerPoint, and PDF files.
No comments:
Post a Comment