• Developed a service to estimate naval ship's docking procedure i.e. analyze the speed and process of docking. Worked on building an end to end pipeline for instance segmentation of corrosion patches in a ship's hull during dry-docking.• Automated the data labeling procedure of images using Roboflow's Autodistill and Auto-labeling. Post feature engineering, I introduced MMSegmentation, an open source tool, for the state of the art segmentation of videos.• Utilized CUDA Kernels using Python and PyTorch, enabling the distributed model training on different gpu servers. Dockerized and deployed the model on AWS.

Worked on Data Security Posture Management System.• Used GoLang to scan PII data from databases• Developed packages to detect metadata on EC2 instances • Handled both relational and directed databases to manage and analyse the personal identity informations stored in them.• Successfully deployed a methodology for scanning PII data across multiple databases using workload scanning. Ensuredcompliance with regulatory requirements by implementing secure protocols and achieving 100% data audit accuracy.

- Responsible for connecting with sponsors to sponsor the fest.- Helped in the proper maintenance and smooth organizing of the Carnival of the fest.

• Architected a fashion recommendation system leveraging deep autoencoders to compress high-dimensional data, achieving superior latent feature extraction for outfit suggestions, attaining 19% increase in user base.• Engineered an end-to-end pipeline utilizing PyTorch for autoencoder development, using deep neural networks and batch normalization to optimize convergence and avoid overfitting, resulting in a 30% improvement in model results.• Utilized autoencoders for dimensionality reduction, reducing the input space from 10,000+ feature vectors to a compact 128-dimensional latent space, accelerating training time by 40% while preserving high recommendation accuracy.