CS 203: Software Tools and Techniques for AI

Lecture Slides & Lab Exercises

Course Information

Instructor: Prof. Nipun Batra, IIT Gandhinagar

Semester: January 2026

Format: 1.5-hour lecture + 3-hour hands-on lab each week

Course Website: nipunbatra.github.io/stt-ai-26

Repository: github.com/nipunbatra/stt-ai-teaching

About This Course

This 15-week course covers the complete software engineering stack for modern AI development. While most ML courses focus on algorithms, this course focuses on the engineering practices that make ML systems work in the real world.

Learning Outcomes

By the end of this course, you will:

Build end-to-end ML systems — From data collection to production deployment
Choose the right tools — Know which tool to use for any data/ML task
Debug systematically — Understand why things break and fix them
Write production code — Code that others can understand, maintain, and reproduce

What You’ll Learn

Module	Key Skills
Data Engineering	APIs, web scraping, data validation, annotation pipelines
Labeling at Scale	Active learning, weak supervision, LLM-based labeling
Model Development	AutoML, transfer learning, LLM fine-tuning
Deployment	Docker, FastAPI, Streamlit, CI/CD
Production	Model optimization, profiling, drift monitoring

Quick Navigation

Part I — Data Engineering (Weeks 1-5)

Week	Topic	Lecture	Lab
1	Data Collection	PDF ∙ HTML	Notebook
2	Data Validation	PDF ∙ HTML	Notebook
3	Data Labeling	PDF ∙ HTML	Notebook
4	Optimizing Labeling	PDF ∙ HTML	Notebook
5	Data Augmentation	PDF ∙ HTML	Notebook

Part II — Model Development & Deployment (Weeks 6-11)

Week	Topic	Lecture	Lab
6	LLM APIs	PDF ∙ HTML	Notebook
7	Model Development	PDF ∙ HTML	Slides
8	Reproducibility	PDF ∙ HTML	Slides
9	Interactive Demos	PDF ∙ HTML	Slides
10	HTTP & APIs	PDF ∙ HTML	Slides
11	Git & CI/CD	PDF ∙ HTML	Slides

Part III — Production & Optimization (Weeks 12-15)

Week	Topic	Lecture	Lab
12	Edge Deployment	PDF ∙ HTML	Slides
13	Profiling	PDF ∙ HTML	Slides
14	Monitoring	PDF ∙ HTML	Slides
15	Summary	PDF ∙ HTML	—

Part I: Data Engineering (Weeks 1-5)

Learning Goal: Transform raw, messy data into clean, labeled datasets ready for ML training.

Week 1: Data Collection

Build robust data collection pipelines using HTTP APIs and web scraping.

Key Tools: Chrome DevTools, curl, requests, BeautifulSoup, Playwright

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

HTTP fundamentals (methods, status codes, headers)
REST APIs and authentication
Rate limiting and pagination
Web scraping with BeautifulSoup
Building a movie data collector

Week 2: Data Validation & Quality

Ensure data integrity through validation schemas and cleaning pipelines.

Key Tools: jq, csvkit, Pydantic, pandas

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

CLI data inspection (head, tail, wc, jq)
Schema validation with JSON Schema and Pydantic
Data profiling and quality metrics
Handling missing values and type mismatches

Week 3: Data Labeling & Annotation

Set up annotation workflows and measure inter-annotator agreement.

Key Tools: Label Studio, Cohen’s Kappa, CVAT

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

Annotation task types (text, image, audio, video)
Setting up Label Studio
Inter-annotator agreement metrics (Kappa, IoU)
Quality control and annotation guidelines

Week 4: Optimizing Labeling

Reduce labeling costs using active learning, weak supervision, and LLM-based labeling.

Key Tools: modAL, Snorkel, cleanlab, Gemini API

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

Active learning strategies (uncertainty sampling, QBC)
Weak supervision with labeling functions (Snorkel)
LLM-based labeling with Gemini API
Handling noisy labels (cleanlab)

Week 5: Data Augmentation

Expand training datasets through synthetic transformations.

Key Tools: Albumentations, nlpaug, audiomentations

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

Image augmentation (geometric, color, noise)
Text augmentation (synonym replacement, back-translation)
Audio augmentation (noise, pitch, speed)
AutoAugment and test-time augmentation

Part II: Model Development & Deployment (Weeks 6-11)

Learning Goal: Build, train, and deploy ML models as production-ready applications.

Week 6: LLM APIs & Multimodal AI

Leverage foundation models for text, vision, and audio tasks.

Key Tools: Gemini API, OpenRouter, OpenAI API

	Materials
Lecture	PDF ∙ HTML ∙ Source
Lab (Notebook)	View ∙ Download .ipynb

Topics Covered:

LLM fundamentals (tokens, temperature, sampling)
Prompt engineering (zero-shot, few-shot, chain-of-thought)
Multimodal capabilities (vision, audio, video)
Structured outputs and function calling

Week 7: Model Development & Training

Train and evaluate machine learning models effectively.

Key Tools: scikit-learn, AutoGluon, PyTorch, LoRA

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Baseline models and cross-validation
AutoML with AutoGluon
Transfer learning and fine-tuning
LLM fine-tuning with LoRA

Week 8: Reproducibility & Environments

Ensure experiments are reproducible and shareable.

Key Tools: Docker, DVC, MLflow, Weights & Biases

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Virtual environments and dependency management
Docker containers for ML
Data versioning with DVC
Experiment tracking with MLflow

Week 9: Interactive AI Demos

Build and deploy interactive AI applications.

Key Tools: Streamlit, Gradio, Hugging Face Spaces

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Streamlit app development
Gradio interfaces
State management and callbacks
Deployment to Hugging Face Spaces

Week 10: HTTP, APIs & FastAPI

Develop production-grade ML APIs.

Key Tools: FastAPI, curl, Pydantic

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

REST API design principles
FastAPI basics and async programming
Request validation with Pydantic
Serving ML models as APIs

Week 11: Git, GitHub Actions & CI/CD

Automate testing and deployment with CI/CD pipelines.

Key Tools: PyGithub, GitHub Actions, pytest

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Git automation and hooks
GitHub API with PyGithub
CI/CD workflow design
Automated testing for ML

Part III: Production & Optimization (Weeks 12-15)

Learning Goal: Optimize, deploy, and monitor ML systems in production environments.

Week 12: Deployment on Constrained Devices

Optimize models for edge devices and mobile.

Key Tools: ONNX, Quantization, Pruning

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Model compression techniques
Quantization (INT8, FP16)
Pruning and distillation
ONNX Runtime deployment

Week 13: Profiling & Optimization

Identify and fix performance bottlenecks.

Key Tools: PyTorch Profiler, NVIDIA Nsight, AMP

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

GPU profiling and bottleneck analysis
Mixed precision training (AMP)
Knowledge distillation
Memory optimization

Week 14: Model Monitoring & Observability

Detect and respond to model degradation in production.

Key Tools: Evidently AI, Prometheus, Data drift detection

	PDF	HTML	Source
Lecture	Download	View	GitHub
Lab	Download	View	GitHub

Topics Covered:

Data drift vs concept drift
Statistical tests for drift detection
Monitoring dashboards
Alerting and retraining triggers

Week 15: Course Summary & Future Trends

Review the ML pipeline and explore emerging topics.

Key Topics: LLMOps, AI Agents, Edge AI, Final Projects

	PDF	HTML	Source
Lecture	Download	View	GitHub

Topics Covered:

Complete ML pipeline review
Emerging trends (LLMOps, AI Agents)
Career paths in ML engineering
Final project presentations

Additional Resources

Useful Links

Course Syllabus: View on main website
Course Specification: COURSE_SPEC.md - pedagogical philosophy and content guidelines
Build Instructions: See README.md for building slides locally
Report Issues: GitHub Issues
Diagram Generators: diagram-generators/ - Python scripts for all course diagrams

Core Tools Covered

Category	Tools
Data Collection	requests, BeautifulSoup, Playwright
Data Quality	Pydantic, pandas, jq
Labeling	Label Studio, modAL, Snorkel
LLM APIs	Gemini API, OpenAI API
Model Training	scikit-learn, AutoGluon, PyTorch
Deployment	FastAPI, Streamlit, Docker
Testing & CI	pytest, GitHub Actions
Monitoring	Evidently AI, Prometheus

Course Materials by Prof. Nipun Batra | IIT Gandhinagar

Last updated: December 2024