have project on multimodal sentiment analysis of hindi-english memes but i want to increase accuracy from 0.51 to 0.85 atleast

I am looking for a freelancer who can help me increase the accuracy of my multimodal sentiment analysis project on Hindi-English memes. Currently, my sentiment analysis uses both text and images as data and is achieving an accuracy between 50-75% using machine learning algorithms. However, I want to improve this accuracy further.

Ideal Skills and Experience:
- Strong knowledge and experience in multimodal sentiment analysis
- Expertise in working with Hindi-English memes
- Proficiency in machine learning algorithms and techniques
- Ability to optimize and fine-tune existing sentiment analysis models
- Familiarity with deep learning algorithms is a plus
- Excellent problem-solving and analytical skills

If you have the expertise and experience to enhance the accuracy of my sentiment analysis project, please reach out to me.

WORK DONE SO FAR : I have dataset and code prepared for multimodal sentiment analysis of hindi english memes. i have used ROBERTa for text modality and CLIP for image modality. The present accuracy is 0.51 , i want to increase it to be atleast 0.85

MY REQUIRMENT:
The accuracy should be at least 0.85 or more
Visualisations of code(in graphs) are optional
the model should be multimodal i.e for both text and image

NOTE : to be completed in 5-7 days of assigning