This paper describes our team's collaborative efforts in participating in the Track 1 for Speaker Diarization of the Diarization of Speaker and Language in Conversational Environments (DISPLACE) Challenge 2024. Our submission focuses on creating a diarization system that is robust to noisy conditions, as well as high amounts of overlapped speech. We conduct an exhaustive study on each component of a hybrid system using techniques such as semi-supervised learning, ensemble of several systems and experiment with both a neural overlap detection module, as well as a post-processing technique using an external overlap detection system. Our final system achieves a diarization error rate (DER) of 28.04% on Phase 1 Eval set, representing a relative improvement of 19.33% compared to the baseline DER of 34.76%.