AI ENGINEER INTERN

• Developed an automated system using Python, BeautifulSoup, and NLP to process 280+ articles weekly from 40 publications, creating summary reports and reducing manual research time by 15 hours/week, with CI/CD pipelines ensuring reliable daily deployments.
• Developed intelligent outreach system using LLM-powered email generation with PostgreSQL database integration, creating personalized journalist campaigns with 85% draft approval rate and automated review workflows using Flask API endpoints.
• Built trend analysis tool using Google Trends API, RAG, and scikit-learn machine learning algorithms to identify emerging topics and predict PR opportunities, within 7-day window achieving 73% accuracy.

AI/ML RESEARCH INTERN

• Developed preprocessing pipeline using Python, OpenCV, and NumPy on 1,200+ fundus images (CLAHE, denoising, vessel segmentation) to enhance diagnostic features.
• Trained CNN model in TensorFlow/PyTorch, achieving 82% accuracy and 0.78 F1-score for early-stage retinopathy detection, reducing manual workload.
• Built automated validation with cross-validation and Grad-CAM explainability, generating heatmaps to support ophthalmologist review.

DEEP LEARNING FOR CRIME PATTERN RECOGNITION, IEEE

January 2023 – May 2023

• Implemented XGBoost, CatBoost, TabNet, hybrid 1D CNN to predict crime patterns using Chicago, Boston & SF datasets.
• Achieved accuracy improvements using feature-oriented datasets, with XGBoost for Chicago 96.51% and San Francisco 26.50%.
• Demonstrated the effectiveness of the Bagging classifier for Boston, achieving the highest accuracy of 20.80%.

DETECTION OF INTOXICATION IN AUTOMOBILE DRIVERS, IEEE

July 2022 – November 2022

• Designed a CNN-based system to detect intoxication through ocular cues, using DL for better accuracy and real-time applicability.
• Evaluated VGG16, VGG19, MobileNet V2, ResNet50, & LSTM+Attention to identify the most effective model.
• Achieved highest validation accuracy with VGG16 86.72%.