AI Engineering : Model Deployment, MLOps & Agentic AI

Build a simple machine learning model with Anaconda and Spyder, then save it as a binary file while focusing on deployment workflows for packaging, serving, and integrating models.

Build a k nearest neighbors classification model using age and salary from store purchase data; train-test split, scale features, and evaluate with a confusion matrix achieving 87.5% accuracy.

Save the KNN classifier and standard scaler in binary format with Python's pickle, enabling predictions in a new environment.

Generate complete python code to train, evaluate, and save the trained model and its scaler in binary format from a dataset using AI tools and well-crafted prompts.

Load classifier and scaler pickle files in Google Colab, fetch via wget from GitHub, predict age and salary in a new environment, and print the probability.

Expose a machine learning model over a rest api using python flask, build a hello world endpoint, handle post requests with json data, and run the service on localhost:8000.

Build a rest api for a machine learning model that returns a prediction via http using age and salary, with a Python rest service and client.

Sign up for a free Google Cloud Platform account and receive a $300 credit for a three-month trial to explore cloud services for machine learning and big data.

Create a Google Cloud VM, enable compute engine API, install flask, numpy, and scikit-learn, deploy a REST API with pickle files, and expose it on 0.0.0.0:8006 for external access.

Delete vm instances when not in use to protect your free credit balance in Google Cloud; use the vm instances page and the three dots to delete, then recreate later.

Build a serverless machine learning API using Google Cloud Functions and a bucket-stored model. Invoke via http for scalable, pay-per-use machine learning predictions.

Create a rest api in Google Colab using Flask Ngrok to expose a model with a pickle classifier and scaler, enabling a public endpoint for predictions for quick prototyping.

Build a FastAPI rest endpoint for your model with ChatGPT assistance. Explore deployment options across cloud platforms, compare costs and trade-offs, and generate sample code with GPT and related tools.

Install and use the Postman rest client to sign in with Gmail, send post requests to a ml model api with json payloads (age, salary), and verify status 200.

MLOps blends data scientists and operations teams to productionize machine learning with CI/CD, continuous training, data versioning, and model monitoring—leveraging MLflow and deployment tools.

Explore MLflow, an open source platform from Databricks that manages the machine learning lifecycle—from experiment tracking and deployment to a central model registry with project, model, and workflow capabilities.

Learn MLflow tracking, an API and UI for logging parameters, metrics, code versions, and artifacts across experiments and runs to visualize, compare, and organize model training.

Install MLflow in Windows and Mac using conda and pip, verify the MLflow UI at localhost:5000, and import MLflow in a Python program to run hands-on labs.

Enable mlflow auto logging to automatically capture parameters, metrics, and artifacts for scikit-learn models like knn, without explicit tracking, and enable deployment via the mlflow rest api.

explore how to use MLflow programmatically with Python to fetch tracking information, including the tracking URI, experiment name by ID, experiment by name, and run ID, via the REST API.

Register and version machine learning models with the MLflow Model Registry, linking each version to exact experiments, parameters, metrics, and artifacts for reproducible production workflows.

Explore enterprise AI engineering on Databricks by building, tracking, registering, and deploying ML models with MLflow, using notebooks and the built-in generative AI assistant.

Register the best-performing logistic regression model from Databricks experiments in the workspace catalog using Unity Catalog, manage versions automatically, and prepare for deployment in the next lecture.

Deploy registered models with Databricks managed serving, create serving endpoint for real-time predictions, test via rest api, Python, curl, or SQL, and monitor metrics while updating versions without endpoint changes.

Discover how to build, deploy, and test models using natural language conversations. Leverage generative AI like ChatGPT to speed up development beyond manual coding.

Install Visual Studio Code, set up GitHub Copilot in agent mode for vibe coding, link your GitHub account, and use Copilot chat to deploy ML models.

Create and deploy a rest api for an ml model using a simple GitHub Copilot prompt, generating code, libraries, documentation, and tests, then validate json predictions and error handling.

Deploy a serverless machine learning API on AWS by using S3 for model binaries, Lambda for inference, and API Gateway for a REST endpoint; leverage Copilot for deployment.

Automate model deployment with GitHub actions workflows to retrain on new data, validate performance, and deploy updates to cloud endpoints or serverless APIs, ensuring up-to-date, tested ML models on AWS.

Explore the history and architecture of large language models, from transformer foundations to the GPT series, and their capabilities, limitations, and other transformer models like Bert, Lamda, Llama, and Xlnet.

Build a simple large language model from scratch using Hugging Face and PyTorch in Google Colab, focusing on next-word prediction with a transformer architecture.

Learn to invoke OpenAI text and image generation models from Python, installing the OpenAI package, creating an API key, and running GPT-3.5 turbo chat and DALL-E style image generation.

Develop a Python-based chatbot by leveraging the OpenAI API and ChatGPT model, building a Colab notebook, setting up the client, and implementing a looping conversation with display of history.

Build your first AI agent using the OpenAI Python SDK in Google Colab, configure the OpenAI client with a hidden API key, and define system prompts to guide behavior.

Deploy an ai agent as a rest API using FastAPI, uvicorn, and the OpenAI SDK, exposed via ngrok for testing, with Pydantic validation and a /agent endpoint.

Build an AI agent capable of web searches and live information retrieval to answer questions with up-to-date data, using a reusable research function and web tool reasoning.

Build an AI agent with memory to maintain context across turns by storing conversations in a Python list, enabling contextual understanding and natural, persistent dialogue, supporting real-world applications.

Build a complete multi-turn ai chatbot by managing conversation history with a messages list and a system prompt, using the OpenAI SDK to generate context-aware responses.

Build a multi-tool AI agent with the OpenAI SDK and Python, selecting weather, stock prices, a calculator, and web search based on user requests.

Enable tracing to visualize an AI agent’s complete execution with the OpenAI Agents SDK, inspecting model invocations, prompts, responses, timing, and metadata in the OpenAI dashboard.

Build a multi-agent artificial intelligence stock analyst using the OpenAI Agents SDK, with three specialized agents for price analysis, news analysis, and investment summary, plus automated email delivery.

Set up a Google ADK development environment using Cloud Code as an AI coding assistant, install Google ADK with pip, and generate a Gemini API key from Google AI Studio.

Migrate an OpenAI Agents SDK-based multi-agent stock monitoring project to Google ADK, preserving architecture while demonstrating framework migration through prompts from ChatGPT and Claude Code.