IIT Delhi Certificate Programme in Applied Data Science and Artificial Intelligence: From Fundamentals to Deployment - Batch 04

Module 3: Optimising Data for Machine Learning Models

• Data Cleaning Techniques
• Data Normalisation & Standardisation
• Feature Selection
• Dimensionality Reduction
• Handling Categorical Variables
• Feature Engineering
• Balancing Datasets
• Include SQL for Dataset Querying

Learning Outcomes

• Clean and preprocess raw datasets by handling missing values and outliers.
• Normalize and standardize data for consistent model input.
• Apply feature selection and dimensionality reduction techniques.
• Encode categorical variables and engineer new features.
• Balance imbalanced datasets to improve model fairness.
• Use basic SQL queries to extract, filter, and join data from structured databases.

Module 4: Machine Learning with Examples

• Supervised Learning (Linear, Logistic Regression)
• Classification Algorithms (Decision Trees, KNN, Naive Bayes, SVM)
• Ensemble Methods (Random Forest, Gradient Boosting)
• Clustering Algorithms (K-means, DBSCAN)
• Model Evaluation Metrics
• Cross-Validation
• Hyperparameter Tuning
• XGBoost, LightGBM, Stacking/Blending
• Focused mini-project: "Credit Risk Classifier using ML"

Learning Outcomes

• Build supervised models for regression and classification tasks.
• Implement popular ML algorithms like Decision Trees, SVM, KNN, and Naive Bayes.
• Use ensemble methods including Random Forest and Gradient Boosting.
• Apply advanced models like XGBoost and LightGBM for high performance.
• Combine models using stacking and blending for better accuracy.
  1. Evaluate models using metrics like accuracy, F1-score, and ROC-AUC.
  2. Tune models using cross-validation and hyperparameter search.
• Apply concepts in a real-world mini-project: Credit Risk Classifier.

Module 5: Deep Learning

• Neural Networks Basics
• Training Neural Networks
• Convolutional Neural Networks (CNNs)
• Recurrent Neural Networks (RNNs)
• Transfer Learning (e.g., ResNet, BERT, Transformers)
• Hands-On with TensorFlow/Keras in Google Colab

Learning Outcomes

• Understand the architecture and training of neural networks
• Build and train CNNs for image tasks and RNNs for sequence data
• Apply Transfer Learning using pre-trained models like ResNet (vision) and BERT (text)
• Develop and train deep learning models using TensorFlow/Keras in Google Colab
  1. Gain hands-on experience in building scalable, real-world DL solutions

Module 6: Applied Industry Cases

• OpenCV:
  1. Image processing,
  2. face detection,
  3. contour analysis,
  4. object tracking,
  5. real-time video apps.
• NLP
  1. Generative AI and LLMs
  2. Prompt engineering
  3. Using OpenAI APIs
  4. Building Q&A bots with LLMs
  5. Langchain Framework
  6. Dspy
• AI project
  1. AI in Healthcare
  2. Forecasting using Time Series
  3. E-commerce Recommender System

Learning Outcomes

• Apply OpenCV for image processing, face detection, object tracking, and real-time video analysis.
• Use NLP and LLMs for prompt engineering, Q&A bots, and text generation with OpenAI APIs.
• Build smart applications using LangChain, Dspy, and Generative AI techniques.
• Develop AI solutions for healthcare, time series forecasting, and recommender systems.
• Gain hands-on experience in building real-world AI/ML projects across multiple domains.

Module 7: Applied model Deployment and Special Topics

• Introduction to Model Deployment
• Containerisation with Docker
• Deployment Frameworks (Flask, FastAPI)
• Cloud Deployment (AWS, GCP, Azure)
• Model Monitoring and Management
• CI/CD for ML Models
• Detecting and diagnosing faults
• MLOps Principles
• MLflow or W&B for model tracking
• Real deployment demo (e.g., Streamlit app + backend API)

Learning Outcomes

• Understand the end-to-end process of deploying machine learning models in production.
• Containerise machine learning models using Docker for scalable deployment.
• Deploy models as APIs using frameworks such as Flask and FastAPI.
• Implement cloud-based deployment solutions using AWS, GCP, or Azure.
• Monitor model performance in production and manage updates to deployed models.
• Integrate CI/CD pipelines for continuous model deployment and scaling using MLflow or W&B.
• Apply MLOps principles to manage the entire machine learning lifecycle from development to deployment.
• Build real-world apps with Streamlit and backend APIs.

Self Paced Content offered by Timepro:

• MLOps Content ( USP ) Async by Timespro
• Gen AI LLM Development Async by Timespro