LLMs and AI Agents for Business

Explore technologies used to build first project, including LLM providers such as Duroc, OpenAI, and Hugging Face, Link Chain orchestration, Streamlit front end, Python and Google Colab, and vector databases.

Install and import essential libraries in Google Colab to build an automated content generation assistant, including chain, chain community, chain rock, and ipywidgets, and prepare the user interface.

Build a user interface by adding a generate button and an output display, wiring a click handler to render generated content from a topic, tone, length, audience, hashtags, and keywords.

Connect Google Colab to the LLM using Grok, obtain and securely store the API key, configure model parameters (temperature, max tokens, timeout, retries), and review rate limits and free models.

Format prompts by defining system and human prompts, using templates in Google Colab. Build prompt chains with a chat prompt template and apply output parsers to view results.

Explore prompt engineering by building prompts with role prompting, templates, and chains; use expert or SEO-focused prompts and apply zero-shot, one-shot, and few-shot techniques to guide model and invoke results.

Apply structured prompting to program language model behavior, coding instructions, examples, and constraints to generate topic-specific, audience-focused posts with SEO elements, CTAs, hashtags, and keywords.

Build a function that accepts the lm and prompt to create the template and chain, returning the generated result; wire output forms and a generate button for content creation.

Set up local execution by installing python and visual studio code, opening the project folder, installing libraries with pip, and running streamlit to generate content.

Learn how to run LLMs locally in Google Colab by downloading a model, enabling GPU, and applying bits and bytes quantization from Hugging Face Transformers to run large models efficiently.

Define a customer service project using a smart virtual assistant that answers questions from internal documents via a vector database, reducing costs and maintaining clear, secure responses.

Install and configure LangChain, transformers, pypdf, and Hugging Face tools in Google Colab to build a retrieval-augmented chatbot that uses internal documents.

Learn how to load an LLM module in Colab, configure an API key, set the model and temperature, and build a prompt chain to generate password change guidance.

Define the context and prompt for a retrieval augmented generation workflow by creating a context variable and a prompt variable; test how the algorithm uses context to answer password questions.

Implement the Rag pipeline by uploading documents, splitting them into chunks, applying embeddings, and storing them in a vector database for efficient retrieval.

Break large documents into chunks with a recursive character text splitter (chunk size 500, overlap 50) to improve retrieval; use embeddings with Hugging Face and store in a vector store.

Create and manage an in-memory store that assigns session IDs to individual chat histories, and use a history-enabled chain to track human and AI messages within a streamlit app.

Assemble indexing and retrieval steps by loading PDFs, chunking text, generating embeddings with HuggingFace, creating a vector store and retriever, and testing a chain-based chat workflow.

Learn how to build a user-friendly GUI for an LLM-based assistant using Streamlit, including setup, environment configuration, and a chat UI with session state and retrieval-based responses.

An ai-powered hr project that classifies resumes, analyzes alignments with job requirements, and generates structured json outputs, including a compatibility score and autogenerated interview questions.

Upload resumes in Google Colab, extract text from PDFs with PyMuPDF, and prepare AI-generated curriculums to rank candidates by job requirements and suggest personalized interview questions.

Customize prompts and the chain to optimize CV analysis against job descriptions. Build templates, parse structured data, and save results in json to show alignment with requirements.

Add new fields to the CV analysis schema, including strengths, areas for development, and important considerations, then test with process CV to visualize structured results and guide interviews.

Calculate a compatibility score from 0 to 10 for CV against a job. Apply 35% experience, 25% technical skills, 15% education, 15% strengths, and up to 10% penalties for weaknesses.

Finish the streamlit gui by creating session state keys for uploader key and selected cv, enabling pdf uploads and exporting results to json and csv in a two-column layout.

Define an ai-powered, personalized learning project that automatically generates level-based exercises with step-by-step explanations, grounded in reference documents, using rag techniques, delivered via a chat assistant with Google Drive export.

Develop a form-driven workflow to generate personalized exercises by selecting level, topic, quantity, and interests, using IPython widgets, loading a reasoning model like Sikh AR1, and preparing export options.

Develop a function to generate exercises from user parameters, build prompts for the algorithm, and tailor questions and explanations to an intermediate topic with optional interests.

Export exercises to Google Drive from Colab by mounting drive, creating a folder, and generating timestamped file names; export a docx via pandoc and save to drive.

Implement rag steps by loading documents, chunking them into 512–1024 token blocks with 20–30% overlap, and storing them in a vector database for retrieval and generation.

Load real documents into a pipeline by uploading astronomy and biology PDFs. Organize them in an articles directory and split content into markdown-based chunks for embedding and rag-based question generation.

Connect to a cloud vector store, run similarity searches on indexed documents with embeddings, tune results using k and score thresholds, and load a retriever for the collection.

Learn AI agents define goals, analyze steps, call APIs, use calculators, search documents, and access databases, retain context to plan sequences, and adapt as data changes using LangGraph’s controlled workflows.

Install the link-graph library, load a general purpose model, and define the agent state to manage messages. Build and connect start, agent, and end nodes, then compile and visualize graph.

Run the chatbot by feeding user input through a length graph flow and looping to keep the conversation active, printing responses and noting need for advanced tools for complex calculations.

Landgraf integrates external tools to extend llm capabilities beyond language. This lecture guides building a mathematical calculator tool, wiring it into the agent to perform calculations with pi and e.

This lecture adds a memory system to maintain state across interactions, showing memory saver usage, graph updates, and a division and multiplication workflow with a math calculator tool.

Learn to build a search tool and a retrieval tool for context-based information using tavli and a vector database to enable rag in business ai agents.

Launch the finance department project to automate reading of lengthy reports and spreadsheets. Provide automatic summaries and transparent traceability with a natural-language interface using llama index and pandas query engine.

Prepare the finance department project by building an LMS-based AI agent that interprets financial documents (reports and spreadsheets), summarizes contents, and answers questions, from data ingestion to analysis.

Learn to load and parse a monthly financial report from a PDF, using a simple directory reader to load data, inspect pages, and concatenate content for analysis.

Explore the ReAct agent, a reasoning plus acting framework that blends chain of thought with memory and tool use, observe results, and adapt actions in a unified decision loop.

Learn to use the React prompt within a RAG pipeline, turning documents into vectors for semantic search with a vector store index and embedding models.

Create and configure a query engine to retrieve top three semantically similar snippets from a pre indexed data set, then build an agent to answer complex questions from long documents.

Build a Streamlit-based user interface to test AI agents by installing libraries, setting API keys, and creating app code that reads documents, summarizes them, and runs a streaming agent.

Build and test a streamlit GUI for uploading documents, verify file selection, load pdf or csv files, and use a Panda's query engine to summarize content.

Create a two-column streamlit UI with an input and button to send queries to an AI agent, then display responses, agent logs, and document summaries for pdfs or csvs.

Explore multi-agent systems by coordinating specialized agents with clear roles. See how the crew AI framework enables collaborative reasoning, task planning, and tool use to orchestrate production-ready AI solutions.

Install crew AI and set up tools for a travel planner multi-agent system, verify version 0.159.0, load the module with a Google API key, and instantiate the Gemini lm class.

Create and configure specialized agents for travel planning, such as city information, travel logistics, and personal itinerary planners, to gather destination insights and deliver a complete itinerary.

Define and create tasks for AI agents, outlining descriptions, expected outputs, and parameters; link tasks across city info, logistics planning, and itinerary generation to ensure coherent, data-driven trip planning.

Orchestrate a multi-agent crew to execute city info, logistics, and itinerary planning tasks using sequential execution for dependent steps, with generated travel itineraries and markdown outputs.

Add a new agent specializing in communication to the multi-agent system, creating a translated guide with practical phrases and cultural tips based on the itinerary.

Implement the final application with Streamlit, install libraries, and configure API keys; organize code into trip tools, trip components, and trip utils, and build organized agents.

Build a travel itinerary GUI with Streamlit, validating form fields, formatting dates, and updating status. Generate multi-agent results, PDFs and markdown outputs across four tabs for travel reports and logistics.

Define a retail analytics project using LLMs and AI agents to analyze customer reviews with sentiment classification and summarization, enabling natural-language to SQL queries via Lang graph.

Install and import data libraries, load a Hugging Face Amazon product reviews dataset, and select key columns like rating, title, text, and timestamp for preprocessing.

Pre-process data by converting timestamps to readable dates, removing milliseconds, and renaming columns. Filter to the top 25 most reviewed products, then visualize ratings with matplotlib and seaborn.

Connects a language model to a SQLite database by loading Gemini 2.0, setting the Google API key, and visualizing results from the reviews table while replacing stray symbols.

Develop and apply sentiment analysis to customer reviews, classifying overall sentiment as positive, neutral, or negative, and summarize key positives, negatives, and common complaints for product insights.

Transform the logic into a LangGraph agent that uses an inner join between reviews and products to generate queries from a product name input, escaping quotes to prevent sql errors.

Learn how three agent nodes handle customer interaction, generate queries from product state, execute queries with a BigQuery tool, and analyze feedback for sentiment and summaries.

Retrieve the database schema, including structure, column names, and data types, using Lang graph tools. Generate a safe SQLite query that targets product reviews, limits results, and avoids data modification.

Validate SQL queries before execution in an LLM-driven pipeline using a validation prompt to ensure safe, correct queries and robust error handling.

Learn how the nl2sql agent's final output node uses a prompt-based control, query results, and a max-queries safeguard to decide final delivery or further queries.

Build an nl2sql agent 5 workflow by configuring a state graph with nodes and edges (list tables, get schema, generate query, execute query, final answer), then test and visualize flow.

Create a Streamlit GUI to load CSV data, build the database, and manage the Google API key, then run analysis with an agent and log outputs.

Install libraries and import image modules, differentiating Agno and PIL image classes. Upload and visualize medical images in Google Colab, then apply pre-processing for llm input with Gemini models.

Perform pre-processing of the image by resizing to a width of 600 pixels with preserved aspect ratio for the Gemini model, then save to a temporary path using PIL.

Learn to craft prompts for medical image analysis, including concise objective prompts, classification prompts, and in-depth clinical analysis prompts that provide context and terminology.

Empower a medical agent with external research tools to search up-to-date literature and clinical protocols. The session demonstrates integrating Tavli and PubMed tools for medical image analysis research.

Build a complete image analysis pipeline with two agents: one analyzes medical content in an image, the other searches for clinical references and updated evidence.

Implement the user interface with Streamlit to upload a medical image, configure API keys, select models, run an analysis pipeline, and visualize results and related articles.

Explore artificial intelligence and data science with high-quality videos on machine learning, deep learning, computer vision, natural language processing, algorithms, and more, in a monthly-updated course platform offering certificates.