AI Engineer.

github

Chaitanya Surabattuni  • Peerlist

Worked on surveillance with the goal of detecting enemies within predetermined areas in military zones for increased protection. This system efficiently detects possible threats by using Convolutional Neural Networks (CNN) and ML. My in-depth work with OpenCV makes it possible to perform thorough visual analysis via camera-based observation, which improves  capacity to identify and evaluate risks faster to train the models.

Autonomous Aerial Surveillance Using DJI Tell

Developed an automobile vent enhancement system with the main goal of preventing children and animals from being left alone in cars. This method is intended to stop suffocation situations that result in fatalities. By using control sensors, it keeps an eye on the quality of the air at regular intervals and makes timely, automated modifications to keep the inside of the car safe.

Car vent enhancement

Used TensorFlow to predict fraudulent transactions, achieved 79% F-1 score.

Credit Card Fraud Detection

Developed a machine learning model to identify network hazards and cyber attacks, focusing on data balancing and cleaning of logs from compromised networks.

Cyber Attack Detection Using Machine Learning

Developed using PyTorch, a deep reinforcement learning model for search engine ranking, achieving an NDCG@1 score of 0.5075, outperforming leading supervised learning models on the LETOR dataset.

DeepQRank

Used computer vision and natural language processing techniques to convert pixel data into human-readable captions for scene understanding. Leveraged VGG16 model and COCO dataset for training and testing, integrating CNNs and RNNs for caption generation. Tackled issues related to the subjective nature of image captions and ensured model validation for output accuracy. Enhanced accessibility for visually impaired individuals, improved content recognition systems.

Image to Text generator

Translated generic words in sign language to captions in live meetings using mediapipe. Achieved really good accuracies.