Update: I am on the industry and academic job market!

Welcome! :)

I am Nastaran, a final-year Ph.D student at the Max Planck Institute for Software Systems in Kaiserslautern, Germany, where I am advised by Manuel Gomez-Rodriguez. I recently submitted my PhD thesis on “Predictive Accuracy and Fairness in Human-AI Teams,” which contributes to a deeper understanding of opportunities and risks in AI-assisted decision-making.

During my PhD I interned at Meta, where I worked with Niek Tax and Daniel Haimovich on multicalibration of ML models, and I visited Amazon, where I worked with Dominik Janzing on root-cause analysis of outliers from a causal perspective. In 2023, I was selected to participate in the Machine Learning Summer School in Okinawa.

I did my undergrad in computer engineering at Ferdowsi University of Mashhad. During my undergrad I interned at IST Austria, working with Amir Goharshady and Krishnendu Chatterjee. Together with Amir, we won the IEEE best student paper award for our POPL paper on data packing. I also interned at the Institute of Computational Perception in JKU Linz, working with Hamid Eghbal-Zadeh and Gerhard Widmer. In 2017, I was awarded the Singapore International Pre-Graduate Award (SIPGA) to visit A*STAR Bioinformatics Institute in Singapore, where I worked with Cheng Li. During this time, we won the first place in iNTUition 2017, a hackathon competition at NTU Singapore, together with Mahdi Abolfazli. In 2018, I was selected to participate in the Heidelberg Laureate Forum.

Research Interests

The ultimate goal of my research is to ensure that Machine Learning (ML) models are efficient, reliable, and safe for those interacting with or affected by them. My current (broad) research interests include uncertainty quantification in ML models (for example, using techniques such as conformal prediction and calibration), LLM safety alignment, provenance (e.g., watermarking), efficiency (e.g., speculative decoding), and, more recently, social and economic aspects in generative-AI.

In my PhD research, my goal was to design a human-AI synergy that leverages the respective strengths of human and AI while mitigating their respective biases. I demonstrate in a line of work that for optimal joint performance in human-AI teams, we must shift our focus from model-centric optimization to team-aware learning. This perspective has inspired the line of research on human-AI complementarity within the human-centric ML literature. In another line of work, I bring attention to the fact that high-stakes decisions made by human-AI teams impact individuals—deciding who receives opportunities—and reshape the distribution of demographic groups in society over time. I identify potential biases in such decisions and propose strategies to mitigate them.

In my free time, I enjoy playing squash, cycling, running, and calisthenics!

Publications

1. Is Your LLM Overcharging You? Tokenization, Transparency, and Incentives
   Tokenization Workshop at ICML’25 and Information Economics and Large Language Models workshop at EC’25
   Ander Artola, Stratis Tsirtsis, Nastaran Okati, and Manuel Gomez-Rodriguez

2. Measuring Multicalibration
   Ido Guy, Daniel Heimovich, Fridolin Linder, Nastaran Okati, Lorenzo Perini, Niek Tax, Mark Tygert (Alphabetical Author Order)

3. Towards Human-AI Complementarity with Prediction Sets
   Giovanni De Toni, Nastaran Okati, Suhas Thejaswi, Eleni Straitouri, Manuel Gomez-Rodriguez
   NeurIPS 2024

4. On the Within-Group Fairness of Screening Classifiers
   Nastaran Okati, Stratis Tsirtsis, and Manuel Gomez-Rodriguez
   ICML 2023

5. Improving Expert Predictions with Conformal Prediction
   Eleni Straitouri, Lequng Wang, Nastaran Okati, Manuel Gomez-Rodriguez
   ICML 2023

6. Differentiable Learning Under Triage
   Nastaran Okati, Abir De, and Manuel Gomez-Rodriguez
   NeurIPS 2021

7. Classification Under Human Assistance
   Nastaran Okati^*, Abir De^*, Ali Zare-Zadeh and Manuel Gomez-Rodriguez (^*Equal contribution)
   AAAI 2021

8. Efficient Parameterized Algorithms for Data Packing
   Krishnendu Chatterjee, Amir Goharshady, Nastaran Okati, and Andreas Pavlogiannis (Alphabetical Author Order)
   POPL 2019
   IEEE Best Student Paper Award

9. Computational Approaches for Stochastic Shortest Path on Succinct MDPs
   Krishnendu Chatterjee, Hongfei Fu, Amir Goharshady, and Nastaran Okati (Alphabetical Author Order)
   IJCAI 2018

10. Root Cause Analysis of Outliers with Missing Structural Knowledge
    Nastaran Okati, Sergio Hernan Garrido Mejia, William Roy Orchard, Patrick Bloebaum, Dominik Janzing

11. Regression Under Human Assistance
    Abir De, Nastaran Okati, Paramita Koley, Niloy Ganguly, and Manuel Gomez-Rodriguez