Tawfiq AADNANE

Catapult High Level Synthesis Hardware Design C++/SystemC RTL Code Generation Design Compiler (DC) Functional Verification/Simulation QA

Graph neural networks for High Level Synthesis Design Space Exploration: Objective: Develop an innovative approach integrating Graph Neural Networks (GNNs) into Catapult HLS to automate and enhance the design space exploration process, improving efficiency and effectiveness in finding optimal hardware designs. Key Sub-Objectives: Graph Representation: Develop a methodology to represent HLS designs as graphs, capturing relationships and dependencies. GNN Architecture: Design and implement different GNNs architecture(GCN,GraphSAGE,GAT) tailored to these graph representations(graph level embedding), focusing on structural and functional characteristics. Data Creation and Synthesis: Develop and synthesize various HLS designs using Catapult, generating data for training and evaluating the GNN model. Compile a comprehensive dataset with performance metrics such as latency and area. Model Optimization: Test different model configurations to achieve optimum performance in accuracy and generalization. Validation and Evaluation: Evaluate the model on separate validations to ensure robustness and accuracy in predicting new data metrics.

Develop an intelligent solution based on the exploitation of natural language processing (NLP) and machine learning techniques to predict the number of budget items within Moroccan public institutions through the analysis of annual finance laws. By combining NLP-based feature extraction, time series analysis and interactive visualisation, the project aims to improve workforce planning strategies and optimise resource distribution in the public sector.

3D computer vision (Machine Learning, Deep Learning, TensorFlow | pytorch , 3D ... [classification, segmentation, génération des objets] )

Create an application to predict timely loan payments in an online P2P lending market using machine learning models by combining financial and social factors, to classify borrowers as "Good" or "Bad".