Projects
Here, I invite you to explore a collection of projects that have fueled my passion for creativity, problem-solving, and innovation. As an aerospace engineering student, I thrive on turning ideas into reality, and each project showcased here reflects my dedication to pushing the boundaries of what's possible.
Nonlinear Aerodynamic Analysis
Utilizing Nonlinear Vortex Correction Method
In the realm of aircraft design and performance analysis, accurate predictions of lift and drag are crucial for optimizing efficiency. Most aerodynamic models however fail to provide accurate results at high angles of attack, limiting the available flight envelope. To overcome these challenges, I have devised a method that integrates experimental 2D lift and drag data with vortex lattice calculations, extending the prediction range of the traditional vortex lattice method into high angles of attack, while giving high accuracy predictions at a fraction of the computational cost of CFD.
Face Tracking & Gesture Control
Algorithm for Autonomous Drones
This project brings together the exciting worlds of robotics, computer vision, and machine learning, resulting in an interactive companion which can track human faces in its vicinity while responding to hand gestures. At its core, this script leverages trained convolutional neural networks to identify the closest human face and the relative position of the body. The algorithm then utilizes a PID controller to move the closest identified face within the centre of the image while keeping the pixel area of the object within a predefined range. Simultaneously, the drone camera finds the position of the hands relative to the face and sends according commands based on the identified configuration.
Reinforcement Learning
Novel Controller for Quadcopters
In a world where drones navigate through unpredictable environments, maintaining stability becomes a paramount challenge. Embracing this challenge head-on, I designed and developed a reinforcement learning controller for the nimble Crazyflie quadcopter, empowering it to hold its position resolutely despite the presence of random disturbances. At the core of this endeavour lies the powerful realm of reinforcement learning, where the quadcopter learns to make informed decisions through iterative experience and continuous adaptation. The cutting-edge controller harnesses the capabilities of neural networks and a PPO algorithm, allowing the drone to acquire intelligence akin to the way a human normally would.
Portfolio Optimization
Utilizing Forward Mean-Variance Optimization
In this project, I embarked on a captivating journey to construct an optimized portfolio comprising the esteemed stocks of the S&P 500. Leveraging historical data spanning from 2000 to 2020, I employed sophisticated algorithms to identify an optimal allocation of assets that would deliver superior returns while efficiently managing risk. The primary objective was to craft a diversified and high-performing portfolio that could potentially outpace the benchmark S&P 500 index over the subsequent three-year period (2020-2023). As the eagerly anticipated years unfolded, the portfolio managed to deliver an impressive return of 19%, significantly surpassing the S&P 500's return of 14% over the same period.
