VoiceRestore: Flow-Matching Transformers for Universal Speech Restoration

Causal Speech Enhancement Based on a Two-Branch Nested U-Net Architecture Using Self-Supervised Speech Embeddings

Final, award-winning B.Sc. Data Science project (TAU, 2025) with my teammate. Developed Deep Learning models (CNNs, GRUs) for vowel decoding from single-neuron brain signals of epilepsy patients. Done in collaboration with Dr. Ariel Tankus (Ichilov) & Prof. Neta Rabin (TAU). Results show feasibility and per-patient variability in decoding accuracy.