Generative AI Research Assisstant

• Developed a Cycle-GAN model in TensorFlow for unpaired CBCT to CT image translation, trained on a dataset of 8,000 images from 80 patients and validated the resulted high-quality CT images through expert human evaluation
• Distributed training parallelly across two NVIDIA 1080Ti GPU, significantly reducing training time to accelerate prototyping
• Built an evaluation-pipeline utilizing Plastimatch and TotalSegmentor to calculate DICE, Hausdorff and Mean Fiducial separation
• Developing a diffusion model to enhance image quality and eliminate the need for adversarial training

Computer Vision Intern

• Reduced cost by migrating the Wire Bonder’s client application to Ubuntu using Qt C++, optimizing memory and performance over 40% by parallelizing processes with QThreads, significantly reducing execution time
• Trained a Cycle-GAN to produce 200K synthetic 2D data-matrix images from a small unpaired augmented dataset and fine-tuned
• Converted the model to ONNX format and accelerated inference by 60% through TensorRT INT8 quantization

Undergraduate Teaching Assisstant

• Educated a cohort of 30+ students on Python programming concepts achieving a 97% pass rate on lab assignments
• Delivered personalized one-on-one tutoring sessions with 15+ students, leading to a 20% average grade improvement

Application Development Analyst

• Designed and implemented REST APIs using Anypoint Studio, enabling seamless integration between Salesforce, SAP and Oracle.
• Certified Mulesoft developer and TechExpressway merit holder with 10,000 prize amount
• Migrated of legacy on-premises micro-services to cloud Mule 4.0, achieving a 50% reduction in downtime and used Anypoint MQ and Async fetch significantly improving performance. Achieved 100% code coverage with MUnit
• Secured the APIs with OAuth 2.0 and JWT authentication protocols, ensuring encrypted credentials within the properties file
• Automated deployments with CI/CD across Dev, QA, UAT and Prod, utilizing Splunk logs for Test-Driven Development (TDD)

Associate Software Engineer

• Developed low-code/no-code cross-platform hybrid applications using Ionic, Angular, and open-source NPM libraries, streamlining the prototyping process by collaborating with cross-functional Node.js and Java backend teams
• Engineered algorithms to facilitate real-time data synchronization and enforce role-based access control mechanisms
• Boosted performance by 25% through efficient querying and indexing in NoSQL databases like CouchDB and PouchDB
• Instrumental in the development of Timesheet Management application tailored for Shell Petroleum, India, adeptly troubleshooting critical bugs and ensuring prompt issue resolution to uphold client satisfaction

AI/ML, Applied Machine Learning, Pattern Recognition, Image Processing, Probabilistic Robotics

Computer Programming, Mechatronics, Computer Aided Design, Finite Element Analysis