The course is divided into 15 Sessions, and would run for 15 weeks including the Capstone Project Preparation/Discussions. Here is the final course structure:
Session 1: Introduction to MLOps; Best practices and tools for managing, deploying, and maintaining ML models.
Session 2: Docker - I; Creating Docker containers from scratch and understanding core concepts.
Session 3: Docker - II; Introduction to Docker Compose for deploying ML applications.
Session 4: PyTorch Lightning and Project Setup; High-performance training and deployment using PyTorch Lightning.
Session 5: Data Version Control (DVC); Managing ML data and models using DVC.
Session 6: Experiment Tracking; Using Tensorboard, MLFlow, and Hydra for experiment tracking and configuration.
Session 7: HyperParameter Optimization; Techniques using Optuna and Bayesian Optimization.
Session 8: Deployment for Demos - Gradio; Creating and sharing ML model demos using Gradio.
Session 9: AWS CrashCourse; Covering EC2, S3, ECS, ECR, and Fargate for ML model deployment.
Session 10: Model Deployment with FastAPI; Deploying ML models using FastAPI.
Session 11: Deploying CLIP; CLIP-as-a-Service with caching on Redis and gRPC endpoints.
Session 12: Model Deployment on Serverless; Introduction to AWS Lambda and Google Cloud Functions.
Session 13: Model Deployment with TorchServe; Deploying ML models using TorchServe.
Session 14: Model Runtime in Browser; Running ML models in the browser.
Session 15: Model Explainability; XAI techniques such as SHAP and LIME.
Session 16: Kubernetes - I; Introduction to Kubernetes and its key concepts.
Session 17: Kubernetes - II; Monitoring and configuring Kubernetes clusters for ML workloads.
Session 18: Deploying Models on a Cluster; Using KServe for deploying ML models on Kubernetes.
Session 19: CI/CD Pipeline - I; Preprocessing, training, and model packaging in CI/CD pipelines.
Session 20: CI/CD Pipeline - II; Model registration and deployment for staging and production.
Session 21: CI/CD Pipeline - III; Connecting the pipeline to a code repository for auto-training and deployment.
Session 22: Auto Scaling Deployment and Stress Testing; Techniques for auto-scaling and stress testing.
Session 23: Model Monitoring and Alerting; Using Grafana and Prometheus for monitoring and alerting.
Session 24: Fine Tuning and Deploying Stable Diffusion; Using DreamBooth for fine-tuning and deployment.
Session 25: Model Performance Optimization; Techniques such as FP16 CUDA_HALF, INT8 Quantization, CUDA AMP, Triton Server, and TensorRT.
Session 26: Fine Tuning and Deploying LLMs; Using FSDP, PEFT, and LoRA for fine-tuning and deployment.
Session 27: Deploying Chat with LLM; Deploying a chat instruction-tuned LLM for conversation.
Session 28: Connecting LLMs with External Data; Augmenting LLMs with external data using in-context learning and data indexing.
Session 29: Capstone - I; Develop and deploy an end-to-end MLOps pipeline as a final project.
Session 30: Capstone - II; Conclusion of the final project and course.