Leloko James Lepolesa

leloko.AI - My AI clone

Key components of the project:

• leloko.AI is an example of an AI-based assistant/chatbot that responds to customer queries, chatting like a human, allowing clients to ask any questions they want to, and responds professionally with the information that only you gave it.
• Answers questions about my work and possibilities to form collaborations with me if interested in the related works
• Gets information from these very pages that you are going through
• Combines with Gemini models to give you good and logical answers
• Deployed and works real-time 24/7

Find it deployed here: leloko.AI

CV-to-Job Recommender

Key components of the project:

• CV-to-Job Recommender is an example of an AI-based recommender system.
• Deployed and works in real time 24/7; it takes the user's CV and extracts keywords. It combines these keywords with the role applied for and the desired location, and then scrapes the job postings from LinkedIn and Pnet that match the user's CV the most.
• Tech choices: Hugging Face model Mistral AI, Fast API for backend, Requests and BeautifulSoup for web scraping, and HTML and JavaScript for the frontend.

Find it deployed here: CV-to-Job Recommender

Dynamic Electric Vehicle Charging Pricing for Load Balancing in Power Distribution Networks based on Collaborative DDPG Agents

Key components of the project:

• Solves the load peak-shaving, valley-filling, and load balancing across distribution networks problem in the presence of EVs
• Explores deep reinforcement learning problem for collaborative control and optimisation
• Compares the performance of the following agents: DDPG, SAC and PPO
• Show the forecasting of the real-world load data using ML algorithms: polynomial regression, XGBoost, and neural networks

The technical write-up can be found at: document

Github link to code can be found at: repository

Multi-level Dynamic Pricing for Electric Vehicles Charging to Balance Load in Distribution Networks Using Q-Learning

Key components of the project:

• Also solves the load peak-shaving, valley-filling, and load balancing across distribution networks problem in the presence of EVs
• Explores advanced data manipulation and engineering to prepare for applied reinforcement learning
• Uses Q-learning agents and custom logic compute for optimisation and control

Github link to code can be found at: repository

Electricity Theft Detection in Smart Grids Based on Deep Neural Network

Key components of the project:

• Solves electricity theft problem based electricity consumption data of the customers
• Uses novel feature engineering techniques to determine which users steal electricity and which ones do not
• Uses the deep neural network for high-accuracy classification results

The technical write-up can be found at: document

Building a regression model to predict concrete compressive strength using the deep learning Keras library

Key components of the project:

• Utilises Keras library for concrete strength prediction based on the dataset of the quantities of the ingredients required
• Explores advanced data manipulation to prepare for applied regression algorithm
• Uses neural network-based regression

Github link to code can be found at: repository

Waste Products Images Classification Using Transfer Learning

Key components of the project:

• Applies transfer learning using the VGG16 model for image classification
• Prepares and preprocesses image data for a machine learning task
• Fine-tunes a pre-trained model to improve classification accuracy
• Visualizes model predictions on test data

Github link to code can be found at: repository

Satellite Images Classification

Key components of the project:

• Applies transfer learning using the VGG16 model for image classification
• Prepares and preprocesses image data for a machine learning task
• Fine-tunes a pre-trained model to improve classification accuracy
• Visualizes model predictions on test data

Github link to code can be found at: repository