aarushi singh · cs'26 · ml research + engineering
recommender systems & diversity
researching methods to improve diversity, relevance, and ranking stability in recommendation pipelines using llm-driven approaches.
emotion recognition with cnns & transformers
benchmarked cnn and transformer architectures for emotion recognition on fer2013 dataset, analyzing accuracy-efficiency tradeoffs.
matrix factorization enhancements
improved ndcg/mrr metrics on matrix factorization models through novel regularization and training strategies.
current interests
- interpretability and controllability of large models post-training
- robustness techniques for ml models on noisy, real-world data
- lightweight models and optimization for faster inference