I like to train neural nets to see and understand images.
I'm a PhD researcher working on semi-supervised approaches for image classification and object detection, with a particular focus on label noise detection. I use tools like Area Under Margin (AUM) and Dataset Cartography to clean and understand training data. I'm also exploring applications of AI in security and biology.
Previously, I've worked across the full stack — from web development with Angular and React, to cloud infrastructure on Azure, to deep learning research with PyTorch and TensorFlow. I enjoy bridging theory and practice, combining principled ML research with real-world systems.
Outside of research, I'm equally passionate about cooking, singing, dancing, acting, and martial arts. I believe creativity fuels innovation.
- Semi-supervised learning — training with limited labels, leveraging unlabeled data effectively
- Label noise detection — identifying and handling noisy annotations in training datasets
- Object detection — working with architectures like YOLOv11 and Faster R-CNN
- Training dynamics analysis — understanding how models learn through metrics like AUM
- Diffusion models — exploring generative approaches for vision tasks
- Medium — blog posts on ML, research notes, and tutorials
- Google Scholar — publications and citations
- ORCiD — researcher ID
- Stack Overflow — Q&A contributions
- YouTube — video content
Languages: Python, C/C++, Java, TypeScript, JavaScript, OCaml, LaTeX, Bash
ML/DL: PyTorch, TensorFlow, Keras, scikit-learn, Hugging Face, OpenCV, spaCy, NLTK
Data: NumPy, Pandas, Matplotlib, SciPy
Infrastructure: Docker, Git, Azure, CUDA, Firebase, Heroku
Databases: MySQL, MongoDB, Microsoft SQL Server
Web: React, Angular, Node.js
I'm open to collaborating on interdisciplinary research combining machine learning with cybersecurity, theoretical computer science, and biology. I'm also interested in vision-related benchmarking and reproducibility studies with other PhD students.
Feel free to reach out: bharanibala77@gmail.com