Ilker Demirel

ilker demirel

PhD Student · MIT CSAIL · ClinicalML

Cambridge, MA · demirel at mit.edu

about

Hey :) I am a 4th-year Computer Science / Artificial Intelligence PhD student at MIT, advised by David Sontag, and an EWSC PhD fellow at the Broad Institute of MIT and Harvard. Previously, I spent two summers at Apple and Microsoft as a research intern. I received my bachelor's in electrical and electronics engineering from Bilkent University, Türkiye.

research

I am interested in representation learning, large language models (LLMs), sample-efficient learning, causal inference, and continual learning, often applied to healthcare.

LLMs + representation learning + sample-efficiency: How can LLMs help learn and design good representations of data—particularly in complex domains—that enable sample-efficient downstream learning?12

Causal inference + sample-efficiency: How can we integrate large-scale real-world evidence with small-scale experimental data to power valid and efficient causal inference?34567

During my PhD, I have also worked on AI for science, including learning representations for the adaptive immune system,8 and predictive modeling with metabolomic and proteomic data.9 Before my PhD, I worked on reinforcement learning for personalized type-1 diabetes management10 and dabbled in communication networks theory.1112

Always happy to chat about research—the best way to reach me is via email.

selected work

LLMs can construct powerful representations and streamline sample-efficient supervised learning

LLMs can construct powerful representations and streamline sample-efficient supervised learning

I. Demirel, L. Shi, Z. Hussain, D. Sontag

arXiv, 2026

Prediction-powered Causal Inferences

Prediction-powered Causal Inferences

R. Cadei, I. Demirel*, P.D. Bartolomeis*, L. Lindorfer, S. Cremer, C. Schmid, F. Locatello

NeurIPS, 2025 — *equal contribution

Using LLMs for Late Multimodal Sensor Fusion for Activity Recognition

Using LLMs for Late Multimodal Sensor Fusion for Activity Recognition

I. Demirel, K. Thakkar, B. Elizalde, M. Espi Marques, A. Sarathy, Y. Bai, U. Srinivas, J. Xu, S. Ren, J. Narain

NeurIPS, Time Series for Health Workshop, 2025 — Work done during a summer internship at Apple.

Prediction-powered Generalization of Causal Inferences

Prediction-powered Generalization of Causal Inferences

I. Demirel, A. Alaa, A. Philippakis, D. Sontag

ICML, 2024

Benchmarking Observational Studies with Experimental Data under Right-Censoring

Benchmarking Observational Studies with Experimental Data under Right-Censoring

I. Demirel, E.D. Brouwer, Z. Hussain, M. Oberst, A. Philippakis, D. Sontag

AISTATS, 2024

ESCADA: Efficient Safety and Context Aware Dose Allocation for Precision Medicine

ESCADA: Efficient Safety and Context Aware Dose Allocation for Precision Medicine

I. Demirel, A.A. Celik, C. Tekin

NeurIPS, 2022

news