Vardhan Mistry

• Designed and implemented DirtyGAN (PyTorch, VAE, CycleGAN) to generate realistic synthetic data, reducing annotation time by 90% and accelerating ML pipeline prototyping with GUI. • Rapidly iterated segmentation benchmarks (U-Net, DeepLabV3, Mask R-CNN) on synthetic & real datasets, enabling data-driven decision-making for production systems. • Collaborated with cross-functional teams to integrate soiling-detection models into edge pipelines, demonstrating prototype production-grade ADAS camera systems.

• Designed and implemented physics-inspired data augmentation pipelines from research, generating synthetic soiling datasets with high-fidelity ground truth masks to improve ADAS perception systems. • Developed a Python-based GUI tool for interactive and scalable data generation and Collaborated with cross-functional teams to deploy a modular data engine supporting automated image/video generation, enhancing scalability and enabling real-world simulation of rare edge-case scenarios.

• Explored AI and computer vision for process optimization in industrial environments, recommending transformation projects. • Integrated OpenCV-based inspection tools into plant workflows, reducing manual QC time by 30%. • Presented prototype results and feasibility analyses to stakeholders, influencing strategic planning.

Computer Vision, Robotics, Autonomous Driving, Automation

Majors: • Computer Application in Mechanical Engineering • Computer Techniques and Programming Language • Control System • Electrical and Electronics • Industrial Engineering