• Implemented and deployed a conversational AI interface that allows users to interact naturally with the diet coach, ask questions, log meals, and receive instant feedback on the dietary choice.

• Built a sentiment-based music recommendation system using NLP and ML techniques and designed an intuitive React user interface enhancing user experience with mood-aligned song suggestions.

• Achieved 92.4% accuracy in sentiment prediction using a bidirectional LSTM model, while utilizing Spotipy for feature extraction and LightGBM for music classification.

• Architected an innovative MIoT android app that facilitated remote patient monitoring, improving healthcare efficiency by 40% and fostering seamless collaboration amongst a cross-functional team of 5 developers.

• Integrated with NVIDIA Jetson Nano and Raspberry Pi for sensor data, implemented user accounts, and enabled real-time health data visualization.

• Developed an interactive Blackjack game leveraging OpenCV, implementing real-time card detection and recognition to elevate user experience.

• Utilized contour detection and perspective transformation for accurate card extraction and recognition, and template matching for suit and rank identification.

• Engineered final year dissertation project that automates the triaging of incoming software bugs to the appropriate developer, based on bug category, frequency, and severity, facilitating optimized developer workload and faster bug resolution.

• Employed LSTM models, training them with open bug repositories from Firefox, Eclipse, and Thunderbird to enhance the accuracy and reliability of the prediction system.

• Compiled and analysed tweets related to ‘online classes during COVID’ using Twitter Developer Account.

• Extracted and Visualized user tweets and metadata via Twitter API using Tweepy and TextBlob.

• Built a text analytics pipeline for text processing, feature representation, vectorization, classification and evaluation.

• Analysed tweet polarity and sentiments based on user location.

• Implemented and compared the performance of Decision Tree Classifier against Logistic Regression, the latter outperforming resulting in 80.4% accuracy.

• Engineered a classification system using Amazon Fine Food Reviews based on 500,000 user views using NLP and ML.

• Applied various text pre-processing techniques: Tokenization, stemming, lemmatization, and normalization.

• Implemented multiple feature extraction and vectorization methods using TF-IDF, Count Vectorizer, Word2Vec and n-grams, visualized using TSNE

• Analysed positive and negative reviews using topic modelling (Latent Dirichlet Allocation), visualized using pyLDAvis.

• Developed and compared the performance of Linear Regression, Decision Tree Classifier and Multinomial Naïve Bayes against LSTM resulting in 83.4% accuracy

• Worked with a sample of German Street Sign Recognition Dataset (grayscale images)

• Applied various data preparation, correlation, feature selection, probabilistic data analysis and data visualization techniques using Weka and Python

• Implemented multiple classification, clustering algorithms to perform street sign recognition tasks

• Developed and compared the performance of different Decision Tree Classifiers, Linear Classifier, Multi-Layer Perceptron (MLP) and Convolutional Neural Networks (CNN)

• Built a multi-layer Artificial Neural Network (ANN) to approximate mathematical functions

• Experimented with multiple activation functions - Sigmoid, TanH, Cosine, Gaussian

• Trained and optimized the ANN using Particle Swarm Optimization (PSO) algorithm

• Implemented Markov Decision Process and Reinforcement Learning based algorithms that uses adaptive intelligence to play the Tic Tac Toe game using Java