Generative AI with AI Agents & MCP for Developers

Explore how generative AI, powered by large language models, creates text, images, audio, and video from unstructured data and prompts, with real-world apps like ChatGPT, Gemini, and Llama.

Explore large language models, the core of generative ai, which understand and generate text using transformer architecture with encoder and decoder layers, pre-trained via transfer learning, and prompt engineering.

Delve into the transformer architecture, its encoder-decoder stack, self-attention and multi-head attention, and the role of positional encoding and normalization.

Explain how ChatGPT is trained through generative pre-training, supervised fine tuning, and reinforcement learning, including the role of reward models and human feedback in optimizing a ChatGPT model.

Explore the basic architecture of generative AI applications, detailing LLMs, orchestration frameworks, and vector databases, and explain level one versus level two apps with Python backends and optional UI.

Master advanced generative AI application architectures by mapping frontend and backend components, integrating external APIs, cloud storage, and vector databases, and deploying LM ops across level 1 to 3 architectures.

Explore a three-level architecture for llm powered apps: level one with orchestration framework and vector database, level two with a POC UI, level three with external APIs and cloud storage.

Explore a professional level generative AI application demo using Lama index projects such as Sake Insight; covers end-to-end rag-based architecture with document uploads, cloud storage, and OpenAI GPT models.

Evaluate foundation models by criteria—accuracy, cost, latency, and privacy—to select GPT-based OpenAI or open-source options for production-ready or PoC deployments.

Discover the tool stack for gen AI applications, covering orchestration frameworks, vector databases, cloud storage, frontend and backend development, LM providers, and CI/CD with AWS deployment.

Explore orchestration frameworks for scalable ai solutions, including long chain, Lang Chain, Lama Index, and OpenAI, with practical demos and rag concepts for building llm-powered apps.

Explore rag, the retrieval augmented generation approach, enabling large language models to access private data and current information via embedding-based vector databases and retrievers for augmented generation.

Explore the Rag technique for generative AI, covering context window limits, token counts, chunking, embeddings, vector databases, semantic search, and how Rag differs from in-context learning and fine-tuning.

Learn the core components of RAG, including data loader, text splitter, embedding model, vector database, similarity search, retriever, and LLM, plus common implementation challenges.

Address common RAG implementation challenges by optimizing the retriever and similarity search for fast responses, with guidance on selecting an orchestration framework for professional AI-powered applications.

Compare universal orchestration frameworks like long chain and lama index to load diverse models, connect vector databases, and enable memory and multi-agent workflows for production-grade AI apps.

Learn LangChain, the orchestration framework for generative AI, tracing its evolution from completion to chat models, the lcl expression language, and its ecosystem for memory, rag, runnable, and agents.

Learn to connect with leading LLMs, including OpenAI GPT and open-source models like llama and mistral, using Langton and Langston tools, with prompt templates, chains, and output parsers.

Learn how to build and use prompt templates to inject logic into LLM interactions, including completion and chat models, with system and user prompts and few-shot prompts.

Explore chains inside Lantern, learn how executable actions form runnable chains by combining a prompt, a chat model, and an output parser; see Langton expression language pipe syntax.

Learn to reformat large language model outputs using an output parser in LinkedIn, switch between JSON and text formats, and build custom formats with pedantic v1 and prompt templates.

Load custom data from diverse formats with data loaders. Grasp the basic retrieval augmented generation concept and its components: splitter, embedding, vector store, and retrieval, with a practical rag demo.

Learn the rag workflow: split data into chunks, generate embeddings with an embedding model, store vectors in a vector database, and use a retriever to query a large language model.

Explore a basic rag using an expression language, build data splitting, embeddings, and a vector store, then connect a retriever to a large language model for accurate responses.

Discover memory management in LangChain, including temporary and permanent memory, and implement buffer memory and conversation window memory to preserve or limit user context, using the legacy engine.

Learn the long chain expression language (lcl) for building runnable chains, distinguish it from legacy chains, and master left-to-right execution, stream and batch alternatives, and built-in runnable concepts.

Explore built-in runnables in LCEL, including runnable pass through, runnable lambda, and runnable parallel, and learn how to compose them into chains with an item getter to execute parallel operations.

Explore built-in functions in runnable, including dot bind, to add arguments and stop a model when Ronaldo appears within LCL chains. See a rack demo using the LCL expression language.

Learn to build a combine chain by nesting a chain inside another, using prompt templates, a chat model, and an output parser. The lecture shows politician examples and language variations.

Demonstrates a RAG demo loading data from a URL with a web-based loader, extracting post content, title, and header, then storing vectors in chroma db for retrieval.

Explore the LangChain ecosystem, including Lang Smith, Lang Serve, and Lang Graph, to monitor, deploy, and build agents and multi-agent systems with real-time data capabilities.

Explore how Lang serve builds LM apps as REST APIs with a web interface, testable via fast API, Flask, or Streamlit, and sharable for production in a real demo.

Explore building ai agents with LangGraph to create simple and multi-agent systems using llms, tools, and memory, including real-time web search and memory-aware conversations.

Discover how to build, monitor, debug, trace, and evaluate an LM powered application with LM ops using Lang Smith, including cost and latency tracking.