Hire this AI Trainer
Sign in or create an account to invite AI Trainers to your job.
Image
Document
Segmentation
Diagnosis
Text Generation
Fine-tuningI have hands-on experience working with AI training data, annotation workflows, and dataset preparation through multiple machine learning and computer vision projects. My work has involved preprocessing large datasets, identifying and labeling relevant patterns, separating useful and non-useful image features, and improving data quality for model training. In my structural crack detection project, I developed preprocessing pipelines to distinguish crack and non-crack elements using geometric descriptors and trained neural networks on curated datasets for accurate segmentation and classification. I also worked on medical imaging datasets for a CNN-based pneumonia detection system, where I performed image preprocessing, augmentation, validation, and dataset optimization to improve training performance and model reliability. Additionally, I have experience generating and organizing synthetic AI training data using GANs for footwear image generation and applying clustering and classification techniques to improve dataset understanding and categorization. My technical background in Python, TensorFlow, PyTorch, Pandas, NumPy, and Scikit-learn enables me to efficiently manage annotation-related tasks, data cleaning, quality checking, and training data preparation for AI systems. I am comfortable working with both structured and unstructured datasets and have strong attention to detail, consistency, and accuracy, which are essential for high-quality AI training and annotation workflows.
DocumentSegmentationAnalyzed large-scale RTI FIR and iRAD crash datasets to identify pedestrian and two-wheeler risk patterns, spatial hotspots, and severity trends. Converted stakeholder interview data into structured quantitative risk indices and metrics for analytical modeling and decision-making workflows.
ImagePerformed synthetic data generation for footwear images using a GAN-based approach. The project created a set of realistic footwear samples and then used clustering and nearest-neighbor methods to group and classify shoe categories. Generated outputs were evaluated with quality metrics and visualizations to assess diversity and realism. • Implemented a GAN to generate 30+ footwear image samples. • Applied PCA with K-Means and K-Nearest Neighbours for style clustering and category classification. • Evaluated generated outputs using quality metrics and PCA visualization. • Produced diverse synthetic samples for footwear style/category exploration.
DiagnosisCreated a CNN-based pneumonia detection system using chest X-ray inputs. The work focused on training and evaluating an optimized model with dataset preprocessing, augmentation, and regularization for better early detection. Model performance was assessed to reach high classification accuracy. • Used preprocessing and augmentation to improve training data quality. • Applied regularization to reduce overfitting. • Trained an optimized CNN for pneumonia vs. non-pneumonia classification. • Reported achieved accuracy of 89.74% on validation/testing.
SegmentationWorked on an AI pipeline to detect and quantify cracks in structural members using deep learning. The project involved preprocessing to isolate cracks from non-crack elements and training a neural network to classify crack vs. non-crack features for improved segmentation. Outputs were validated through a deep learning framework designed for real-time structural image analysis. • Developed a preprocessing pipeline using geometric filters and descriptors. • Trained a neural network for crack vs. non-crack feature classification. • Designed a deep learning framework targeting crack detection and segmentation. • Enabled real-time segmentation on structural image inputs.
Bachelor of Technology, Civil Engineering
Event Coordinator
Research and Data Analyst Intern
