Professional AI Portfolio

Welcome to my Professional AI Portfolio! This repository contains detailed projects showcasing my work in Machine Learning (ML), Deep Learning (DL), HuggingFace Large Language Models (LLMs), and Computer Vision (CV). Below you will find a summary of each project, the AI books' chapters I've studied, and the MIT Deep Learning basic course I've completed.

Projects

Project 1: Obesity Predictor

• Techniques: ETL (Extract, Transform, Load), Grid Search Cross Validation (GridCV), Mean Squared Error (MSE)
• Algorithms: Logistic Regression, Linear SVC, Random Forest Regression, Linear Regression
• Description: Developed a model to predict obesity based on various health metrics. Applied multiple ML algorithms to find the best performing model.

Project 2: Star Type Classification

• Techniques: ETL, GridCV
• Algorithms: Logistic Regression, Linear SVC, Random Forest Classifier, K-Nearest Neighbors (KNN), Gradient Boosting Classifier
• Description: Created a classification model to categorize stars into different types using astronomical data. Experimented with various ML algorithms to improve accuracy.

Project 3: Hotel KPIs and Critical Dimensions Models

• Techniques: ETL, GridCV
• Algorithms: MultiOutputRegressor with RandomForestRegressor, Single models for KPIs (Lead time with RandomForestRegressor, GradientBoostingRegressor, XGBoost), MultiOutputClassifier with RandomForestClassifier
• Description: Built models to predict key performance indicators (KPIs) for a hotel, such as lead time. Utilized both regression and classification approaches for comprehensive analysis.

Project 4: Apple Quality Classification

• Techniques: Professional ML pipelines, GridCV, Cross Validation, RandomizedSearchCV, Ensemble Methods
• Algorithms: LightGBM, XGBoost
• Description: Implemented a model to classify apple quality. Leveraged advanced techniques such as ensemble methods and fine-tuning to optimize model performance.

Project 5: Apple Clustering for Unsupervised Learning

• Techniques: Data Ingestion
• Algorithms: K-Means Clustering, Hierarchical Clustering, Spectral Clustering
• Description: Performed clustering analysis on apple data to identify patterns and group similar apples together using various clustering algorithms.

Project 6: Clustering Algorithms and Techniques

• Techniques: Data Ingestion
• Algorithms: Agglomerative Clustering, Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Mean Shift Algorithm, K-Means Clustering, Elbow Analysis
• Description: Explored different clustering techniques to group data points effectively. Analyzed results using elbow analysis to determine the optimal number of clusters.

Project 7: Anomaly Detection

• Techniques: Data Engineering Lifecycle, ML Pipelines for Data Processing, Normalization, Standardization, Model Training
• Algorithms: Robust covariance, Isolation Forest, One-Class SVM, Local Outlier Factor
• Description: Developed an anomaly detection system to identify outliers in datasets. Implemented and compared multiple algorithms to ensure robust detection.

Project 8: Anomaly and Novelty Detection

• Techniques: Data Engineering Lifecycle
• Algorithms: Isolation Forest, Local Outlier Factor
• Description: Focused on detecting both anomalies and novel data points using advanced ML techniques. Employed a lifecycle approach to data engineering for comprehensive analysis.

Project 9: Dimensionality Reduction

• Techniques: Multiple ML Pipelines
• Algorithms: Principal Component Analysis (PCA), Kernel PCA, RandomForestClassifier, SoftmaxClassifier
• Description: Applied dimensionality reduction techniques to simplify datasets while retaining important information. Evaluated performance improvements in classification tasks.

Project 10: Data Visualization

• Techniques: ML Pipelines (Imputers, Encoders)
• Algorithms: t-distributed Stochastic Neighbor Embedding (t-SNE), PCA Visualization, PCA + t-SNE, Locally Linear Embedding (LLE), PCA + LLE, Linear Discriminant Analysis (LDA)
• Description: Created visualizations to interpret complex datasets. Combined multiple visualization techniques to enhance data understanding and insight.

Project 11: Association Rule Learning

• Techniques: Reusable Models with Functions
• Algorithms: Multi-Layer Perceptron (MLP), Random Forest Classifier, Gradient Boosting Classifier, XGBoost, Logistic Classifier, Decision Tree Classifier, Linear SVC, SVC, K-Nearest Neighbors, Naive Bayes, Ada Boost Classifier, LightGBM Classifier, Linear Discriminant Analysis, Quadratic Discriminant Analysis, Gaussian Process Classifier
• Algorithms: Apriori Algorithm, Eclat Algorithm
• Description: Implemented association rule learning to discover interesting relationships between variables in large datasets. Used a variety of classification algorithms to improve rule discovery.

Project 12: Semi-Supervised Learning Classification

• Techniques: Google AI Human-Centered Design Approach
• Algorithms: Logistic Regression, Label Propagation, Label Spreading, Pseudo-Labelling
• Description: Developed classification models using semi-supervised learning techniques. Applied Google’s human-centered design principles to enhance model usability and performance.

Project 13: Semi-Supervised Learning Regression

• Techniques: Google AI Human-Centered Design Approach
• Algorithms: Pseudo-Labelling
• Description: Focused on regression tasks using semi-supervised learning. Leveraged pseudo-labelling to enhance model accuracy with limited labelled data.

Project 14: Introduction to Deep Reinforcement Learning

• Techniques: Markov Decision Processes (MDPs)
• Algorithms: QLearningAgent, Convolutional Neural Networks (CNNs)
• Datasets: MNIST
• Description: Introduced the fundamentals of deep reinforcement learning using MDPs. Implemented Q-learning agents and CNN architectures for hands-on practice.

Project 15: CNN with CIFAR-10 Dataset

• Techniques: CNN Architectures
• Datasets: CIFAR-10
• Description: Trained and fine-tuned CNN models on the CIFAR-10 dataset. Focused on best practices and techniques to achieve high accuracy in image classification.

Project 16: CNN with CIFAR-100 Dataset

• Techniques: CNN Architectures
• Datasets: CIFAR-100
• Description: Extended CNN training to the CIFAR-100 dataset. Addressed challenges of classifying a larger number of categories and fine-tuned models accordingly.

Project 17: CNN with CIFAR-10, Best Practices and Real Applications for Image Detection

• Techniques: CNN Architectures
• Datasets: CIFAR-10
• Description: Applied CNNs to real-world image detection tasks. Emphasized best practices and advanced techniques for practical applications.

Project 18: RNN with HuggingFace Transformers

• Applications: Sentiment Analysis, Speech Recognizer
• Description: Implemented Recurrent Neural Networks (RNNs) using HuggingFace Transformers. Developed models for sentiment analysis and speech recognition tasks.

Project 19: Advanced CV with Document Image Classification

• Applications: Document Parsing, Document Visual Question Answering (DocVQA)
• Description: Explored advanced computer vision techniques for document image classification. Implemented models for document parsing and visual question answering.

AI Books' Chapters

CB001 - Hands on Machine Learning Book Chapters

1. Introduction to ML
   • Overview of ML, types of ML systems, main challenges of ML.
2. End-to-End ML Project
   • Example of a real-life ML project, data handling, feature engineering.
3. Classification
   • Techniques and algorithms for classification tasks, performance measurement.
4. Training Models
   • Linear models, polynomial regression, learning curves.
5. Support Vector Machines
   • SVMs for linear and nonlinear classification, regression, outlier detection.
6. Decision Trees
   • Decision trees, random forests, feature importance.
7. Ensemble Learning and Random Forests
   • Bagging, boosting, stacking, random forests.
8. Dimensionality Reduction
   • PCA, Kernel PCA, LLE, t-SNE.
9. Unsupervised Learning Techniques
   • Clustering, DBSCAN, Gaussian Mixtures.
10. Artificial Neural Networks with TensorFlow and Keras
    • Fundamentals of neural networks, training deep neural networks.
11. Training Deep Neural Networks
    • Vanishing/exploding gradients, regularization techniques.
12. Custom Models and Training with TensorFlow
    • Customizing models, training loops, using TensorFlow.
13. Loading and Preprocessing Data with TensorFlow
    • Data pipelines, TFRecord format, data augmentation.
14. Deep Computer Vision Using Convolutional Neural Networks
    • CNNs for image classification, object detection, semantic segmentation.
15. Processing Sequences Using RNNs and CNNs
    • RNNs, LSTMs, GRUs, time series forecasting, NLP.
16. NLP with RNNs and Attention Mechanisms
    • Text generation, machine translation, transformers, BERT model.
17. Representation Learning and Generative Learning
    • Autoencoders, GANs, representation learning techniques.
18. Reinforcement Learning
    • RL fundamentals, Q-learning, policy gradients, deep Q-networks.
19. Training and Deploying TensorFlow Models at Scale
    • Model deployment, TensorFlow Serving, TensorFlow Lite.

MIT Deep Learning Basic Course

1. Deep Learning Basics
   • Introduction to deep learning concepts and terminologies.
2. Introduction to DL
   • Neural networks, binary classification, logistic regression.
3. Computation Graph, Vectorization, Shallow Neural Networks
   • Building computation graphs, vectorized implementations, shallow neural networks from scratch, random initialization.
4. Deep L-layer Neural Networks
   • Implementing deep neural networks with multiple layers using Fashion MNIST dataset.
5. Calculus for Deep Learning (Limits and Intermediate Value Theorem)
   • Fundamental calculus concepts essential for understanding gradient descent and optimization.
6. Calculus for Deep Learning (Differentiation and Derivatives)
   • Techniques for differentiation, chain rule, partial derivatives.
7. Math with Python and Numpy
   • Arrays, L1 and L2 normalization, min-max scaling, normalization techniques using Python and Numpy.
8. Logistic Neural Networks
   • Logistic regression models, neural network implementation for logistic tasks.
9. Shallow Neural Networks
   • Architecture and training of shallow neural networks.
10. Gradient Descent Types
    • Batch gradient descent, stochastic gradient descent, mini-batch gradient descent.
11. Exponentially Weighted Averages, Momentum, RMSProp
    • Techniques to accelerate gradient descent using momentum, RMSProp.
12. Adam, Learning Rate Schedule, GridCV
    • Adam optimizer, learning rate schedules, hyperparameter tuning with GridCV.
13. Pandas, Caviar, Batch Normalization
    • Data manipulation with Pandas, Caviar dataset, batch normalization for training deep networks.
14. Softmax
    • Softmax activation function for multiclass classification.
15. Gradient Checking
    • Verifying the correctness of backpropagation implementations using gradient checking.