I am a Machine Learning Engineer specializing in computer vision and deep learning, currently pursuing my MS in Computer Science at Oregon State University. My work focuses on developing efficient object detection and segmentation models, automating data labeling pipelines, and optimizing models for real-time deployment.
- 🎓 MS in Computer Science, Oregon State University (2023–2025)
- 💼 Jr. Machine Learning Engineer, Microtec Inc
- 📚 Published researcher in edge computing & deep learning
- Architected automated labeling system combining object detectors + SAM, boosting labeling speed by 50%.
- Built assisted annotation tool with PySide6 + SAM backend, doubling labeling throughput and replacing legacy tools in production.
- Trained foundational object detector (300k+ images, 20 classes) achieving 87% mAP and implemented continual learning (≤3% forgetting).
- Developed novel loss functions & pretraining methods, reducing FP instances by 80% and cutting data requirements by 30%.
- Masters’ Thesis: Developed image reconstruction attack pipeline using DiffBIR + ControlNet, improving SSIM by 73% over GAN baseline.
- Conv-MAE Pretraining: Designed MAE-inspired masked pretraining for hybrid conv-transformer (QT-ViT), boosting ImageNet-100 accuracy to 80.18%.
- Languages: Python, C++, SQL
- ML/DL Frameworks: PyTorch, CUDA, ONNX, TensorRT, OpenMMLab, FFCV
- Tools: Git, Docker, Azure DevOps, Jenkins, MLflow, Valohai
- Other: PySide6, OpenCV, NumPy, Pytest
- Inference at the Edge for Complex Deep Learning Applications with Multiple Models and Accelerators – ICCCNT 2023
- 📧 Email: kuppusaa@oregonstate.edu
- 🔗 LinkedIn: Ashwanth Kuppusamy
- 💻 GitHub: ashwanth-07