Mastering Generative AI: From Python to NLP, GPT 4 & LLMs

Explore how large language models power user-facing generative AI applications, from input prompts to GPU-accelerated processing in cloud data centers, through scalable, multi-layer application architectures.

discover how large language models are built from data, cleaning, and training, using Azure AI Machine Learning Studio and automated NLP tasks such as classification and named entity recognition.

Trace the history of artificial intelligence, examine today’s state and events driving breakneck artificial intelligence innovation, and outline how large language models rely on massive data.

Explore how large language models operate behind the scenes and their impact on generative AI. Assess benefits, risks, and limitations, and learn how foundation models enable future applications.

Explore how tokenizers convert text into numeric chunks and encoding, then see how large language models treat these tokens as a probability machine to predict the next token.

Foundation models are large trained bases using massive text data to start AI apps. Fine-tuning and filtering tailor them for coding assistants, customer support, and sales bots.

Explain how large language models work, their components and logic for generating text, images, or audio, discuss limitations and repurpose foundation models you can fine-tune.

Explore how ChatGPT and OpenAI launched generative AI, clarify terms and models like GPT-3.5 and GPT-4, and note Azure OpenAI language models with paid subscriptions.

Hugging Face serves as a platform and repository to host open source models and spaces, highlighting licenses such as Apache 2.0 and models like Mistral AI for text generation.

Explore Azure AI Studio to deploy large language models in the cloud, access interactive API and playground, and manage prompts with prompt flow for flexible cloud-based AI solutions.

Explore cloud-based solutions and learn to run large language models locally to save costs, then scale to cloud services like Azure while exploring open source models on hugging face space.

Master prompt engineering to improve interactions with large language models by learning zero shot prompting and few shot prompting, adding context, asking concise questions, and evaluating responses for better results.

Explore prompt engineering as a key communication skill for interacting with large language models, using context-rich prompts to elicit accurate, actionable responses.

Master zero-shot, one-shot, and few-shot prompting to guide a large language model with or without examples; learn crafting prompts for tasks like book reviews and Python testing.

Provide clear context when prompting a large language model to generate a Readme for a Python project. Tailor instructions for developers and emphasize Python-specific structure and sections.

Leverage one-shot and few-shot prompt examples to guide JSON-formatted outputs for APIs, transforming city temperature data in Portugal into structured formats.

Discover how to prompt large language models in a chat interface, using few-shot prompts and richer context with examples to improve response quality.

Discover how to craft prompts with tone and persona, refine initial responses from large language models, and use feedback while recognizing limits from excessive context.

Refine prompts with evolving context to tailor responses, switch units, simplify data structures, and explore formats like yaml, xml, and csv for clearer outputs.

Explore advanced prompting techniques to manage context limits, avoid overload, break tasks into smaller steps, use chain of thought, and experiment with beginning, end prompts and to-do lists.

Explore the context limitations of large language models as you paste large data, port code across languages, generate tests, and observe how context length constrains multi-language tasks.

Break down a large task into small chunks to guide a language model in extracting notes from wine csv, filter by year 2015 or newer, and format results as json.

Learn chain of thought prompting to elicit step-by-step reasoning from a large language model, illustrated by an athlete injury case and factors like overtraining, warm-up, and muscle imbalance.

Apply prompt techniques by delivering instructions first, then pasting the code, with up front and end context to improve results. Generate five pytest unit tests for the code.

Explore the limits of context in large language models, practice step-by-step reasoning and smaller tasks, and use front and end prompt cues to elicit clearer, better responses.

Explore how to build generative ai applications by connecting to large language models, including api-based apps, embedded models, and multimodal systems, while examining benefits, challenges, and common pitfalls.

Explore real world generative AI applications, from text-based chat interfaces and customer service bots to image, audio, and email generation, and see how daily tasks gain efficiency through LLMs.

Explore API based applications using HTTP APIs and remote large language models to process prompts and respond, with Python and Rust examples highlighting separation of concerns for generative AI.

Embed a model inside an application using an http API, as shown with a GPT-2 example downloaded and hosted by uvicorn, highlighting internal model containment.

Assess the benefits and risks of large language models through a ChatGPT demo that summarizes bios, highlights privacy and potential inaccuracies, emphasizing their probabilistic nature and the need for verification.

Explore common types of generative applications and architectures, including multimodal options. Learn how lightweight, smaller models and api-based offloading to remote models with gpu resource demands affect performance and deployment.

learn how retrieval-augmented generation lets large language models access fresh data through embeddings and a search database, improving answers by integrating new information with your own system.

Explore retrieval-augmented generation by using vectorized documents to ground and enhance large language models with up-to-date, relevant search results through a vector database and Azure AI search.

Manage data for a retrieval augmented generation app by loading a wine csv with name, region, variety, rating, notes using pandas, converting to records for embeddings in a vector database.

Encode data with sentence transformers to generate embeddings, build an in-memory vector database, and perform cosine-distance search for retrieval-augmented generation (RAG) with LLMs.

Build and query a vector database of wines, import and compress data, and fuse it with an llm via prompts to retrieve a Malbec recommendation from Argentina.

Explore retrieval augmented generation with a large language model, using csv data, pruning and massaging it to create embeddings and a vector database for stronger chat-based responses without retraining.

Deploy a cloud based generative ai application with an http api on Azure, configure cloud components and deployment automation, then verify that the APIs are publicly available and interconnected.

Build a decoupled retrieval augmented generation app on Azure using the Azure OpenAI service and cognitive search, with LangChain, embeddings, and a single /ask endpoint.

Configure Azure container apps and deployment pipelines, align GitHub workflows and container registries, and deploy an application that connects to Azure OpenAI service.

deploys automation using GitHub Actions to build docker containers and deploy to Azure Container Apps, configure ingress port 8000, monitor logs and environment variables, and scale replicas.

Deploy an http api based generative ai app to Azure cloud, using automation and GitHub actions, then verify deployment and understand the required components for end-to-end setup.

Gain a solid foundation in generative AI and master prompt engineering, retrieval augmented generation, and cloud deployment to ship production-ready AI applications and empower leadership decisions to solve big problems.

Explore how to process natural language with Python, from tokenization and vectorization (TF-IDF, word vectors) to sentiment analysis, topic modeling, named entities, and mapping text to external knowledge bases.

Learn to work with unicode in Python 2 and Python 3, covering utf-8 encoding, decoding, and common pitfalls like unicode encode errors.

Explore tokenization and sentence segmentation using NLTK and spaCy to convert text into sentences and tokens, and learn how punctuation, stop words, and regular expressions influence boundaries.

Filter out stop words to reveal the semantic gist of text using English and Spanish stop word lists, tokenization, and a Python set for fast membership checks.

Explore vector representations of documents and words, use tf-idf with idf to measure similarity, and experiment with word vectors from Google and your own.

Discover word vectors, dense 300-length representations that capture word similarity. Learn how skip-gram training predicts surrounding words within a window to reveal analogies like Italy and Rome.

Explore Google word vectors and Google News vectors loaded via Gymnasium, compute cosine similarity, and examine word relationships from mass to mouse and house, plus Freebase vectors.

Learn to train domain-specific word vectors with Jensen, tokenize data, and build document vectors using tag documents, evaluating similarities and analogies with real datasets like 20 newsgroups and tf-idf comparisons.

Explore text structure through parsing with dependency passing, identifying the head word, and applying named entity recognition and part-of-speech tagging using tokenization and Spacey.

Identify the head of a sentence, usually the verb, to reveal the central structure and subject and direct object relationships through dependency.

Identify named entities and apply the IAB format to determine entity types in text, such as geopolitical entities and languages, with practical examples like New York City and French.

Explore sentiment analysis and how text reveals writer feelings, covering machine learning over text, a sentiment dictionary, intensifiers, and negations to capture aspects and cinema analysis.

Explore how Vader sentiment analyzes informal text from social media, including emoticons and slang, using a valence aware dictionary to compute positive, negative, neutral, and compound scores.