Melissa Chehade Email & Phone Number

Beirut, Lebanon

Beirut, Lebanon

Byblos - Lebanon

Munich, Bavaria, Germany

• Researched, summarized and presented state-of-the-art papers related to traffic prediction and multi-agent path deconfliction.• Assembled a neural network architecture inspired from the state-of-the art papers.• Prepared a baselines implementation for the previous traffic prediction models such as HA, ARIMA, VAR, SVR, DCRNN and STG2SEQ.• Deployed state-of-the-art networks using different deep learning frameworks - such as PyTorch, MxNet and Tensorflow - and tested them using different datasets.• Improved available models' results by pre-training them on their provided datasets, adding transformers, optimising parameters and adding LSTM to emphasise on the temporal embeddings.• Used PyTorch geometric’s Graph Attention layers (GAT) to replace Graph Convolutional layers (GCN) to emphasise on the spatial factors.• Visualized clusters of data based on sensors’ location, roads, cities, counties andpostcodes.• Integrated backbones such as ResNet101 and DenseNet into the created network.

Bachelor Of Engineering - Be, Computer Engineering, 3.73/4.0

Master'S Degree, Computer Science