Home
Aritra Bandyopadhyay
Exploring mechanistic interpretability, multimodal systems, and information retrieval. This Quarto template uses an ethereal, soft-gradient glassmorphism aesthetic for a calm, modern portfolio experience.
Research Interests
I’m fascinated by the internal structure of language models, practical information retrieval, and medical-facing computer vision systems. This setup embraces fluid, colorful gradients and frosted glass elements.
Interpretability
Mechanistic interpretability, post-training behavior, and making neural representations transparent.
Engineering
Bridging the gap between research fellowships, startup ML engineering, and reliable deployment.
Frameworks
Maintaining content parity across Quarto, Astro, Jekyll, and Hugo to find the perfect stack.
Featured Work
Differential Evolution Algorithm Based Hyper-Parameter Selection of CNNs for Speech Command Recognition
An evolutionary hyper-parameter search paper exploring the optimal architecture for convolutional neural networks in audio processing.
Contributions
PyTorch
The deep learning stack that keeps showing up in research code and production prototypes.