Comment by _yahya_

Location: Morocco

Remote: Yes

Willing to relocate: Yes

Technologies:

Python, PyTorch, Hugging Face Transformers, Deep Learning, Vision Transformers, Multimodal Large Language Models (MLLMs), Efficient Fine-Tuning

Résumé/CV – Master’s Research Project (2025)

- Developed a parameter-efficient adaptation strategy for frozen MLLMs by learning pixel-space visual prompts that steer the model without modifying its weights.

- Designed and executed experiments on multiple public vision datasets using PyTorch, Hugging Face Transformers, and custom vision-language integration pipelines.

- Achieved performance competitive with full fine-tuning while updating less than 0.001% of total model parameters (e.g., 52K vs. 7B), demonstrating strong generalization and efficiency.

- Paper and code available soon.

Email: c.benmahane@gmail.com