MLOps Training in Perungudi

• Advanced Python programming and Object-Oriented Design for ML
• Writing unit and integration tests using PyTest and mocking
• Code quality enforcement using pre-commit hooks, Black, and Ruff
• Packaging ML projects using Poetry and managing virtual environments
• Implementing structured logging and error handling for production
• Version control best practices and collaborative Git workflows

Beginner Build a tested, packaged, and linted Python ML library with automated CI checks.

• Building optimized, multi-stage Docker images for ML models
• Managing multi-container applications with Docker Compose
• Understanding Kubernetes architecture, nodes, and control planes
• Deploying ML workloads using Pods, Deployments, and Services
• Managing configuration and secrets in Kubernetes
• Introduction to Helm charts for packaging ML applications

Intermediate Dockerize and deploy an ML model API to a local Kubernetes cluster with auto-scaling.

• Managing cloud storage (AWS S3 / GCP GCS) and data lakes
• Orchestrating data workflows using Apache Airflow basics
• Transforming data for ML using dbt (data build tool)
• Handling data quality checks and schema validation at ingestion
• Implementing data partitioning and optimization for ML training
• Managing cloud IAM, security, and access policies for data

Intermediate Build an Airflow DAG extracting raw data, transforming via dbt, and loading to a cloud warehouse.

• Setting up and configuring MLflow for experiment tracking
• Logging parameters, metrics, and artifacts automatically
• Managing the Model Registry, staging, and production transitions
• Using Weights & Biases (W&B) for deep learning experiment tracking
• Tracking model lineage and data-to-model dependencies
• Integrating hyperparameter tuning tools (Optuna) with tracking

Intermediate Deploy an MLflow server to track, compare, and register multiple classification model versions with metadata.

• Introduction to Data Version Control (DVC) for large datasets
• Configuring DVC with remote cloud storage backends
• Understanding the architecture and need for Feature Stores
• Implementing Feast for managing offline and online feature stores
• Ensuring point-in-time correctness for training and serving
• Handling feature transformations and monitoring feature drift

Advanced Build a Feast feature store serving consistent, low-latency features to batch training pipelines and real-time APIs.

• Building CI/CD pipelines for ML using GitHub Actions
• Implementing Continuous Integration (code, data, and model tests)
• Implementing Continuous Delivery (automated model packaging)
• Setting up Continuous Training (CT) triggers based on data drift
• Using CML (Continuous Machine Learning) for PR reporting
• Automating model validation and performance regression testing

Advanced Create a GitHub Actions workflow automatically testing, retraining, and deploying models when new data is pushed.

• Architecting production-grade LLM applications
• Deploying and managing Vector Databases (Pinecone, Milvus)
• Optimizing chunking, embedding, and retrieval strategies
• Implementing RAG evaluation metrics using RAGAS
• Managing prompt versioning and prompt engineering workflows
• Deploying LangChain and LlamaIndex applications at scale

Advanced Build an enterprise knowledge-base chatbot with automated evaluation pipelines measuring retrieval accuracy and hallucination rates.

• Building robust ML APIs using FastAPI and Pydantic
• Scaling inference using Ray Serve for distributed computing
• Deploying deep learning models with NVIDIA Triton Inference Server
• Serving models on Kubernetes using KServe and Seldon
• Implementing dynamic batching and request queuing strategies
• Executing A/B testing, shadow deployments, and canary releases

Advanced Serve a computer vision model using Ray Serve and KServe, configured to auto-scale based on real-time traffic.

• Detecting data drift and concept drift in production models
• Building monitoring dashboards using Evidently AI
• Tracing LLM inputs, outputs, and tool calls using LangSmith/Arize
• Implementing custom business metrics and SLA logging
• Setting up automated alerting systems (PagerDuty, Slack)
• Performing root cause analysis for model performance degradation

Advanced Create a monitoring system detecting data drift in tabular models and tracing LLM hallucinations in GenAI apps.

• Implementing Model Risk Management (MRM) frameworks
• Configuring Role-Based Access Control (RBAC) for ML pipelines
• Understanding adversarial ML and securing model endpoints
• Implementing ML FinOps to optimize GPU and cloud compute costs
• Maintaining audit logs and ensuring compliance (GDPR, SOC2)
• Managing model lineage and automated governance policies

Expert Configure strict access controls, GPU usage alerts, and compliance audit reports for a production ML system.

• Orchestrating complex ML workflows with Kubeflow Pipelines
• Managing scalable workflows using Argo Workflows
• Introduction to ML platform backends (e.g., Zenon, Backdrop)
• Building Internal Developer Portals (Backstage) for ML teams
• Creating self-service infrastructure for data scientists
• Managing multi-tenant ML environments and resource quotas

Expert Build a self-service portal allowing data scientists to spin up isolated training environments and register models.

• Gathering business requirements and designing system architecture
• Building end-to-end automated data and training pipelines
• Deploying models with high-availability and auto-scaling
• Implementing full observability, drift detection, and alerting
• Applying security, governance, and cost-optimization policies
• Creating technical documentation and presenting to stakeholders

Expert Deploy a fully automated, monitored, and secure Intelligent Document Processing (IDP) system from scratch to production.