Multivariate Predictive Maintenance 

In my capacity as a team lead for the multivariate predictive maintenance project at Sand Technologies, our team has made significant strides in developing advanced machine learning models to predict equipment failures proactively. Leveraging a wealth of data sources including sensor data, maintenance logs, and error reports, we meticulously preprocessed the data, conducted comprehensive exploratory data analysis, and engineered relevant features. Our efforts have yielded promising results, with our models demonstrating strong predictive performance. Our focus was on implementing the predictive model using Streamlit, a critical component in deploying the solution for real-time decision-making. While challenges persist, particularly in optimizing the predictive capabilities of the Streamlit app, we remained dedicated to refining our approach and delivering impactful solutions for predictive maintenance at Sand Technologies. We successfully built an app called MaintX, which helps technicians to exactly know how to maintain their equipment. In addition to the application, we also built a dashboard that goes in depth, providing information such as sensory data and the number of machines performing well and/or not.

Movie Recommender System

At ExploreAI Academy, I had the opportunity to contribute to a project centered on building a movie recommendation system using advanced machine learning techniques. In today’s technology-driven world, recommender systems play a crucial role in guiding individuals to make informed choices, particularly in selecting content from an extensive library of options. Our challenge was to develop an algorithm capable of accurately predicting how a user would rate a movie they had not yet viewed, based on their historical preferences. Leveraging content-based or collaborative filtering approaches, we engineered sophisticated algorithms to analyze user behavior and movie attributes, ultimately providing personalized recommendations. This project not only deepened my understanding of machine learning principles but also underscored the importance of data-driven decision-making in enhancing user experiences.

Twitter Sentiment Analysis

In this project, we explored Natural Language Processing (NLP) techniques to analyze and classify public sentiment from Twitter data. The goal was to build a machine learning pipeline that could accurately determine whether a tweet expressed a positive, negative, or neutral sentiment , a valuable tool for businesses, political analysts, and social researchers.The process began with cleaning and preprocessing the text data using techniques such as tokenization, stopword removal, and lemmatization. We then converted the tweets into numerical features using TF-IDF vectorization. After thorough exploratory data analysis, we trained and evaluated multiple machine learning models including Support Vector Classifier (SVC), Bernoulli Naive Bayes, and K-Nearest Neighbors. Model performance was assessed using accuracy scores and runtime efficiency. As shown in the figure below, SVC emerged as the top-performing model with the highest accuracy, though it had the longest runtime. Bernoulli Naive Bayes and KNN models, while faster, demonstrated slightly lower predictive performance.

South African Language Hackathon

I participated in a  hackathon challenge focused on language identification using natural language processing (NLP) techniques. Tasked with identifying the language of text written in any of South Africa's 11 official languages, I leveraged machine learning algorithms and NLP methods to accomplish this goal. By meticulously preprocessing the text data and engineering relevant features, I trained models to accurately classify the language of the text. The success of our approach was measured using the F1-score metric, which evaluates the precision and recall of our models. This project not only honed my skills in NLP and machine learning but also provided valuable experience in tackling real-world language processing tasks.

Employee Attrition 

In my project on employee attrition prediction, I focused on developing a model to identify factors contributing to employee turnover. This project involved detailed exploratory data analysis (EDA), feature engineering, and model evaluation. While I am still working on hyperparameter tuning, the insights gained from this project have significantly enhanced my understanding of predictive modeling and its application in HR analytics. Through this project, I demonstrated my ability to preprocess data, select relevant features, and build a predictive model to address real-world challenges.