Ashish Kumar

Experience

• 

  Iwane Laboratories, Ltd.

  Dec 2020 - Jul 2022

  Key accountabilities:▪ Developed a static C++ library that will contribute to Autonomous driving AI project.▪ Conducted Outlier/Anomaly detection on the in-house dataset with wrong annotation/labels to reduce features for training and reduced time spent to build model by 40%.▪ Performed regularization by using Data Augmentation to decrease the imbalance in the dataset which decreased overfitting by 60%.▪ Integrated feature scaling by normalization that increased accuracy by 10%.▪ Implemented YOLOv4 algorithm for Object detection and CNN techniques like One-shot learning, Transfer learning and normal CNN with 2 pooling techniques like Max Pooling and global Average Pooling for classification.▪ Built the architecture for transfer learning model with varying pre-trained model architectures, optimized and hyper-tuned to get the best performance with least resources, got overall best accuracy of 96.4% and 94% for resource as low as 50 images per class.▪ Transformed the trained model into C++ library by using OpenCV for YOLO model and TensorFlow C API for CNN model to improve the time performance by 50%.▪ Executed Kalman filter algorithm in C++ for Object Tracking across frames increasing the accuracy of object detection by 8%. Show less Developed an desktop application tool for viewing 3D models. The tool enables user to view and interact with any 3D models. The user can change the viewing angle and distance, change the speed of animation, modify the lighting source position direction and color of lights. The application supports all 3D model formats supported by Assimp library. The tool cuts down the time spent for testing models before incorporating into software product by 60%.Below are the areas having experience on:▪ Built desktop application from front to back end using MVVM architecture.▪ Executed multipage WPF application and package installer using WIX toolset.▪ Developed multi-threading codes and handled deadlocks using mutex to improve performance by 40%. ▪ Debugged memory leak and missing DLL, used GLog for logging making debugging process 80% faster.▪ Composed custom events, event handlers, properties, delegates and listeners.▪ Implemented advanced OpenGL concepts like camera, texturing, lighting, shadow mapping, UV mapping, face culling, anti-aliasing using linear algebra and 3D geometry. Show less

  • Software Developer - Computer Vision

    Jul 2021 - Jul 2022

  • Software Developer - Windows

    Dec 2020 - Jun 2021
• 

  HL Klemove India Tech Center (ITC)

  Sept 2022 - now

  Engineer