Batuhan Koyuncu
Batuhan Koyuncu

PhD Candidate & AI Scientist

Saarland University
ELLIS PhD Program
Intor AI

About Me

I build reliable, interpretable, and scalable AI systems for real-world deployment. I am a Ph.D. candidate in Computer Science at Saarland University, part of the ELLIS PhD program, co-advised by Prof. Isabel Valera and Prof. Ole Winther.

My research focuses on probabilistic and generative models for temporal and heterogeneous data, with applications spanning macroeconomic forecasting, healthcare, and audio. I am particularly interested in combining foundation models with agentic AI frameworks to enable robust reasoning, forecasting, and decision-making under uncertainty.

Previously, I was a visiting researcher at the Bank for International Settlements where I built macroeconomic time-series foundation models for central-bank-grade policy simulations. I also work as an AI Scientist at Intor AI, building RAG systems and AI-powered applications.

Download CV
Interests
  • Time-Series Foundation Models
  • Generative AI
  • Probabilistic Modeling
  • Agentic AI Systems
  • LLM Applications & RAG
Education
  • PhD Computer Science

    Saarland University

  • MSc Computer Engineering

    Boğaziçi University

  • BSc Physics

    Boğaziçi University

Recent Publications
(2026). BISTRO: A Foundational Model for Unconditional and Conditional Forecasting of Macroeconomic Time Series. BIS Working Papers, No 1337.
(2025). Temporal Variational Implicit Neural Representations. Under review..
(2025). Hyper-Transforming Latent Diffusion Models. In ICML'25.
(2024). E-ProTran: Efficient Probabilistic Transformers for Forecasting. In Workshop on Structured Probabilistic Inference & Generative Modeling. ICML'24.
(2024). From Laboratory to Everyday Life: Personalized Stress Prediction via Smartwatches. In Machine Learning for Life and Material Science, ML4LMS Workshop. ICML'24.