- 🎓 Bachelor's in Artificial Intelligence
- 🧬 Deeply interested in Natural Language Processing (NLP), Computer Vision, and applied Deep Learning
Assistive BCI system decoding EEG signals into text and speech for individuals with speech impairments.
- Multi-stage pipeline: signal preprocessing → feature extraction → sequence encoding → context modeling → language refinement → TTS
- Two-stage text enhancement (rule-based + GPT-4o-mini)
- BLEU-4: 36.64% | ROUGE-1 F1: 52.76% | BERTScore F1: 90.52% on 2,404 test samples
PyTorchBART-largeFastAPIReact/ViteMongoDB
Surveillance video anomaly detection on the DCSASS dataset with a two-stage pipeline: X3D-S for binary anomaly gating and VideoMAE ViT-B for 13-class crime classification.
- 99.16% weighted test accuracy
- Full-stack app:
FastAPI+React/Vite, dark forensic amber theme - Solved real deployment challenges: Decord for H.264 decoding on Windows, PyTorch Nightly + CUDA 12.8 for RTX 5060 Ti (Blackwell sm_120), YAML config with Pydantic
Fine-tuned Jais-256m via QLoRA on the AHQAD dataset for accurate Arabic medical Q&A.
- Advanced Arabic text normalization, optimized memory usage
- Evaluated on BLEU, ROUGE, and BERTScore
QLoRAJais-256mTransformersNLP
- 📧 eyad.alatifi@gmail.com
- 💼 LinkedIn: Eyad Alatifi
Thanks for stopping by — always open to collaborating on AI, CV, and BCI projects!