Project : A Multilingual Approach to Meme-Based Hate And Offensive Speech Detection in Low Resource Indian Languages. Key Responsibilities & Achievements: Conducted research on Natural Language Processing (NLP), Computer Vision, and Deep Learning techniques for offensive meme detection. Processed and analyzed datasets in Bangla, Hindi, Gujarati, and Bodo including text extraction using OCR. Built and trained multimodal models combining text embeddings (Word2Vec, GloVe, mBERT) and image features (CNN/ResNet). Designed a fusion-based classification system to detect categories like Sentiment, Sarcasm, Vulgarity, Abuse, and Targeted Hate Speech. Performed evaluation using Accuracy, Precision, Recall, F1-score, and Confusion Matrices across multiple languages. Published structured findings with a focus on AI for Social Good and low-resource language processing. Skills gained: Deep Learning • Natural Language Processing (NLP) • Computer Vision • Multimodal Machine Learning • OCR (Tesseract, EasyOCR) • Data Preprocessing • Research Writing • Python • TensorFlow / PyTorch