Projects
ACME Delivery Service
ACME Delivery Service is a web application that allows
individuals to order online food, beverages, etc. The user can
search for the desired store, create an order containing items
only from the same store and retrieve all placed orders. Also
the user can view the list of the most famous stores in general,
per category and the 10 most popular products.
This project involved collaboration and utilized both Angular
and Spring Boot technologies.
Front-End
SkimLit - NLP
This project combines supervised & transfer learning and
utilizes the PubMed 200k RCT dataset and NLP techniques to
classify medical abstract sentences into roles (e.g., objective,
methods, results). It explores various text classification model
architectures and strategies.
It involves preprocessing the PubMed RCT200k dataset, conducting
various modeling experiments, and building a multimodal model to
replicate the architecture proposed in the referenced paper.
Finally we choose the best-performing model for our test data.
The models we developed are the following: Global Average Model,
Pre-trained Embedding Model, Conv1D Character Embedding Model,
Token and Character Hybrid Model, Positional Token Character
Embedding Model, Modified Trihybrid Model with Callbacks
Food Vision - Computer Vision
The "Food Vision Big™" project combines supervised & transfer
learning and aims to surpass DeepFood paper's top-1 accuracy of
77.4% using the Food101 dataset and TensorFlow Datasets.
- The project covers data handling, efficient training, model
building, fine-tuning, and result visualization.
- The model is trained, achieving an accuracy of 83.79%.
- Fine-tuning and evaluation on test data showcase the model's
success in surpassing the targeted accuracy.
- Visualizations include a confusion matrix, F1 scores plot, and
analysis of most wrong predictions.
Mobile Manipulation Capstone
The capstone project involves writing software to plan a
trajectory for the end-effector of a youBot mobile manipulator,
perform odometry as the chassis moves, and perform feedback
control to drive the youBot to pick up a block at a specified
location, carry it to a desired location, and put it down. The
final output is a CSV text file that specifies the
configurations of the chassis, the angles of the four wheels,
and the state of the gripper (open or closed) as a function of
time. The software was tested on the CoppeliaSim simulator.
The project was implemented using Python as part of the online
course Modern Robotics: Mechanics, Planning, and Control
Specialization.
Bachelor & Master Thesis
We used MATLAB and Python to implement theoretical models and numerically calculate parameters, for Matrix representation, functions & tensor analysis and 1-D, 2-D & contour plots.
Bachelor ThesisMaster thesis