Anik Mallick

LLM-agnostic prompt management — typed YAML loading, structured output schema, composable builder API, markdown skills, and a CLI.

No project description available.

Fine-tuned RoBERTa classifier for detecting normative statements in text, with a feature-engineering LR ensemble (deontic keyword scoring, FAISS retrieval, sentence-level aggregation) and an interactive Streamlit inference app.

No project description available.

Benchmarking MLP, CNN, Depthwise Separable CNN, fine-tuned ResNet-18, and MobileNetV2 on CIFAR-10 — exploring accuracy, efficiency, and spatial robustness across architectures.

This project performs a systematic study of different machine learning approaches for handwritten digit classification using the MNIST dataset. The goal is not only to achieve high accuracy, but to understand how different modeling approaches behave on image data.