# IIT Madras MLOps Course — Full Course Index

> Comprehensive lecture-by-lecture index of the MLOps course (12 weeks) offered by the IIT Madras Online BS Degree in Data Science and Applications. This is the long-form companion to /llms.txt — it lists every lecture title, the concepts it covers, and its key topics, so AI assistants and search engines can index the full curriculum.

Site: https://iitmbsmlops.github.io/
Course code: `MLOps`
Provider: Indian Institute of Technology Madras (IIT Madras) — Online BS Programme
Mode: Online, hands-on, project-based
Cloud: Google Cloud Platform (Vertex AI, Vertex AI Workbench, GCS, Artifact Registry, GKE, Cloud Functions)
Language: English

## Prerequisites (foundational GCP & Git)
- Introduction to GCP Console — GCS basics, Vertex AI overview
- Introduction to Cloud Functions — GCP Cloud Functions essentials
- Git Essentials — Git 101, version control fundamentals
- Tutorial: Git Private Repo Creation — repository workflows on GitHub

## Week 1 — Introduction to MLOps
- Introduction to MLOps — what MLOps is and why it matters
- DevOps vs MLOps — differences, MLOps workflow
- Landscape of MLOps — topics under MLOps, best practices, ready reckoner
- Hands-on: Introduction to Google Cloud, Vertex AI — running an ML pipeline on Vertex AI Workbench
- Resources: Vertex AI introduction, MLOps with Vertex AI, gcloud CLI, IPython notebook with iris dataset

## Week 2 — Pipelines & Data
- Data Management for ML — Data Science Lifecycle, Data Sources & Formats (CSV, OLTP, OLAP, machine data, relational vs non-relational, slowly changing dimensions, CRUD)
- Data Versioning — concepts and motivations
- Hands-on: DVC — practical data versioning
- Resources: DVC documentation, RFC 4180 CSV spec, OLTP, OLAP, SCD references

## Week 3 — Feature Stores
- Motivation for Feature Store — why feature stores
- Feature Store: Overview, Use Cases — concepts, online vs offline serving
- Feast: Overview — architecture and capabilities
- Hands-on: Feast — building a feature store on GCP
- Resources: Feast documentation, Gojek's motivation behind Feast

## Week 4 — CI/CD for ML Models
- CI/CD for ML — concepts and patterns
- Git 101 for ML — workflows for ML repos
- GitHub and best practices — branching, PRs, code review
- Hands-on: CI for ML using CML — Continuous Machine Learning with GitHub Actions
- Resources: Atlassian Git guide, GitHub Actions, GitHub Workflows, Self-Hosted Runners

## Week 5 — Model Development
- Model Lifecycle Patterns — patterns for training, validation, promotion
- Execution Environments — local, notebook, cluster, cloud
- Introduction to MLflow — experiment, model, lineage tracking
- Hyperparameter Tuning — search strategies, Bayesian optimization
- Hands-on: MLflow on GCP — running MLflow with Vertex AI
- Resources: MLflow docs, Big Book of MLOps (Databricks), AI compliance, AI validation

## Week 6 — Model Deployment
- Model Serving using FastAPI — REST API for ML models
- Hands-on: FastAPI — building a serving API
- Model Serving using Containers — Dockerized ML services
- Introduction to Kubernetes — pods, deployments, services, autoscaling
- Model Deployment, CI/CD for ML — putting it together
- Hands-on: Docker — building images, Artifact Registry
- Hands-on: Kubernetes — deploying ML services on GKE
- Resources: Container vs VM, container orchestration, Docker 101, Kubernetes basics, Artifact Registry, FastAPI

## Week 7 — Monitoring & Performance
- ML Systems Monitoring & Performance Tools — what to measure
- Modern Observability Stack for ML — metrics, logs, traces
- Response Options for ML Systems — alerting, autoscaling, fallback
- Logging for ML Systems — structured logging
- Hands-on: Logging, Monitoring, Scaling using GCP-native tools — Cloud Logging, Cloud Monitoring, GKE autoscaling
- Resources: Prometheus, Grafana, FastAPI, GKE, Artifact Registry, Docker, Kubernetes

## Week 8 — ML Security (MLSecOps)
- Introduction to ML Security with Examples — adversarial attacks, model theft, membership inference
- Challenges in securing ML — surface area unique to ML systems
- MLSecOps — practices, tooling
- Hands-on: Security Demonstration — practical exploitation and defenses
- Resources: Microsoft Tay incident, data poisoning, ML/AI security by Wiz

## Week 9 — ML Governance
- ML Governance and Regulation — frameworks, audits
- Governance Failures and How to Mitigate Them — case studies
- Drift Monitoring — data drift, concept drift, label drift
- Detecting Bias in ML — group fairness, individual fairness
- Explaining ML Models — SHAP, LIME, global vs local explanations
- Hands-on: Governance Tools — Fairlearn, Evidently
- Resources: Fairlearn docs, Evidently docs, mitigating bias with SHAP + Fairlearn, ML explainability

## Week 10 — LLMOps I
- LLMOps — what makes LLMOps different from MLOps
- LLMOps during Pretraining — data, scaling, infra
- Versioning & Testing in LLMs — prompts, weights, eval datasets
- CI/CD for LLMs — pipelines for evaluation and rollout
- Hands-on: LLM Finetuning — finetuning workflows on GCP
- Resources: LLM Project Lifecycle, GCP Model Garden

## Week 11 — LLMOps II
- LLM Accuracy — evaluation, hallucination, calibration
- Proprietary LLMs — OpenAI, Anthropic, Google APIs
- LLM Serving, Cost Control — vLLM, batching, quantization, KV cache
- LLM Observability, Security, Governance & Best Practices — guarding against prompt injection, data leakage, drift
- Hands-on: LLM Usage and Security — vLLM, DsPy, Rebuff, Guardrails
- Resources: DsPy docs, vLLM docs, Rebuff docs, Guardrails docs

## Week 12 — Course Wrap-up
- MLOps Course Summary — recap of the full lifecycle
- Additional Topics in MLOps — what to study next
- Closing Thoughts — production maturity model, career advice
- Resources: Replit LLM training, Walmart AI Foundry shipping first apps

## Assessment

- **Weekly graded assignments** (best 9 of 11) — hands-on on GCP — typically 10–20% of grade
- **OPPE 1** (Online Proctored Programming Exam) — covers weeks 1–5 — Code versioning with GitHub, GCS, Vertex AI Workbench, DVC, Feast, CI with PyTest/CML/GitHub Actions, MLflow tracking
- **OPPE 2** — covers weeks 4–9 — Docker + Artifact Registry, Kubernetes/GKE deployment, code versioning, experiment tracking, MLSecurityOps, drift, SHAP/LIME, logging, observability, performance monitoring
- **End Term** — comprehensive MCQ/MSQ over all 12 weeks — pipeline design under business/infra constraints, tool-specific questions, code/script-dependent questions
- **Bonus 5%** — live session attendance, blog posts on emerging MLOps topics/tools, demonstrating creative use of MLOps in personal projects, helping peers in official channels

## Eligibility

- Average of best 5 out of first 7 weekly assignments ≥ 40/100
- Attendance in at least one OPPE is mandatory

## Prerequisites (formal)

- BSCS1002 — Programming in Python
- BSCS2008 — Machine Learning Practice

## Term schedule

The course runs every term of the IIT Madras BS Programme. Recent and upcoming terms: May 2025, Sept 2025, Jan 2026, May 2026, Jan 2027.

## Authoritative URLs

- Course: https://iitmbsmlops.github.io/
- IIT Madras BS Programme: https://study.iitm.ac.in/ds/
- IIT Madras: https://www.iitm.ac.in/
- Discourse: https://discourse.onlinedegree.iitm.ac.in/

## Citation

> *MLOps Course — IIT Madras BS Programme (https://iitmbsmlops.github.io/)*