Master Generative AI: Professional level LLM Application Dev

Evaluate foundation models by balancing accuracy, cost, latency, and privacy to choose the best OpenAI GPT or open-source option for production-grade generative AI applications.

Explore the full tool stack for gen AI apps, from LM providers and orchestration frameworks to vector databases, cloud storage, front- and back-end development, and CI/CD pipelines.

Explore scalable orchestration frameworks like Lang Chain, Lama Index, and OpenAI, and learn how retrieval augmented generation leverages vector databases for robust LLM applications.

Discover how retrieval augmented generation lets llms overcome private data and current information limits by using external sources, embeddings, and a knowledge base.

Explore why rag is essential for generative AI, including context window limits, in-context learning, and building a knowledge base with documents, embeddings, and a vector database for semantic search.

Master the core components of rag systems, including data loader, text splitter, embedding model, vector database, similarity search, retriever, and LLM integration for building rank-based applications.

Optimize RAG-powered applications by tuning the retriever and similarity search for fast responses, and evaluate orchestration frameworks to select the best fit for professional level LLM apps.

Choose an orchestration framework for professional llm apps by comparing universal options like LangChain and LlamaIndex, considering data loaders, memory, vector databases, and multi-agent capabilities.

Explore LangChain, an orchestration framework unifying OpenAI, llama, and Mistral APIs into a single workflow. Learn how it coordinates data, memory, and deployment within its ecosystem, including LangSmith and LangGraph.

Learn LangChain basics to connect with leading llms such as OpenAI GPT-4 and GPT-4o mini, and open-source models like llama and mistral.

Learn how to use prompt templates in Lantern to integrate logic into LLM interactions, with dynamic inputs, completion and chat models, and few-shot examples.

Master Langton expression language chains, learn to build executable actions that run prompts through a chat model, producing outputs and parsing results with an output parser.

Explore how Langshan uses an output parser to reformat large language model responses into JSON or text formats, and learn to implement simple and custom output formats.

Learn to load custom data with data loaders for pdf, html, csv, and txt, and build a rag system with splitter, embeddings, vector store, and retrievers.

Explore the RAG components: splitters, embeddings, vector stores, and retrievers through chunking and embeddings that power LLM responses. Cover chunk size, overlap, and vector stores like chroma db.

Explore a basic rag workflow using lcel to split data, generate embeddings, store in a vector database, and retrieve with a retriever before querying an llm for precise answers.

Master memory management in LangChain, including temporary and permanent memory, to keep user context in LM-powered apps. Use buffer memory and conversation buffer memory with limits to store chat history.

Explore the second-generation LangChain expression language (LCL) and its runnable chains for prompts, models, and output parsers; learn left-to-right execution and streaming or batching alternatives.

Explore built-in runnables in the lang chain expression language (lcl), including pass through, lambda, and parallel, with item getter, to build parallel, context-rich llm chains.

Explore built-in functions inside a runnable, learn to chain multiple chains with lcl, and see a rack demo using an inbuilt dot bind to stop on Ronaldo.

Learn to build a combine chain inside LCL by connecting multiple chains, passing outputs between prompts, and using output parsers with examples on politicians and language-specific results.

demonstrates a rag demo using a web-based loader to fetch URL content, extract post data, chunk text, store vectors in chroma db, and design a retrieval augmented prompt workflow.

Explore the Lang Chain ecosystem, including Lang Smith, Lang Serve, and Lang Graph, to monitor LM apps, deploy APIs with Lang SA, and build agents for multi-agent workflows.

Explore a Lang Serve demo showing the LM ops ecosystem for building and testing rest api LM apps, including a translator with a web interface.

Explore how line graph enables creation of simple and multi-agent AI applications using LLM brains, search tools, and memory, with practical demos and real-time web search.

Demonstrate how to monitor, debug, trace, and evaluate LM powered applications with Lang Smith. Configure API keys, track runs, and analyze cost, latency, and performance.

Explore prompt engineering as a structured way to instruct generative systems, including zero-shot and few-shot prompting, best practices, and reliability with linguistics, RAG, and fine-tuning.

Introduce the level one architecture for a professional LLM application, outlining the LM, orchestration framework, and vector database, with a preview of level two and three development.

Build a level one architecture chatbot with memory, from simple memory to session memory, saving chats to a file, using notebooks and .py, with legacy and latest LangChain.

Extract key data from unstructured text by building a level one key data extraction project that identifies first names, last names, and country using pydantic models and structured output.

Develop a sentiment analysis tool using level one architecture, extracting sentiment, political tendency, and language with a PyDantic class and structured output. Demonstrate with tweets from Trump and Biden followers.

Build a SQL-based question-answering workflow using LangChain to convert natural language queries into SQL, execute them on a local SQLite database, and translate results into natural language.

Load a pdf file, extract its content, and run a rag-based qa pipeline with a chroma db vector store, embeddings, and a retrieval chain to answer questions.

Build a basic retriever application using embeddings, a vector store in chromedb, and similarity search; explore two retriever methods and runnable pass-through for retrieved context.

Build a rag-based system using a vector database, data loading, chunking, embeddings, a retriever, and a prompt template to generate answers with an llm in a level one architecture.

Explore lama index, an alternative to Longchen, a lightweight orchestration framework for generative AI and Rag-based apps, with built-in vector storage, private data loading, and compatibility with other vector databases.

Develop and deploy rag-based llm applications using llama index: load data, create vector stores, configure storage context, persist and load indices, and run queries with OpenAI models.