Mehmet Arda Eren

• Built an active-learning framework for supervised regression that uses uncertainty estimates to select the most informative samples — matching full-dataset accuracy with 23% fewer labeled samples, validated in a PyBullet physics environment. • Extended it to cut required labels by a further 42%, while making the selection algorithm lighter and generalizable to any regression-based deep learning model. • Work led to an accepted IEEE conference paper (SII 2026) and a journal manuscript in preparation.

Designed and built agent-based simulations in AnyLogic to evaluate emergent system behavior under varying operational conditions. Translated raw simulation outputs into decision-ready performance metrics, bridging system modeling and practical operational insight.

Research Fellow at Symbiotic Intelligent Systems Research Center (SISReC) with my thesis advisor Prof. Erhan Oztop and Minoru Asada. Research topics are efficient robot learning, uncertainty quantification on deep neural network models

Application of various Machine Learning models on MIMIC-IV, with the supervision of Mehmet Tan and Batuhan Bardak