My research bridges reinforcement learning and human modeling; specifically, how learning agents can improve on their decision making via interacting with a human and receiving feedback in various forms. I am interested in both the theoretical foundations of RL and human modeling leading to novel algorithms with the goal of building systems where human-AI interaction is principled and efficient in terms of improving on sample complexity and regret bounds.

Structured Bandits with Human Assistance Active Mar 2025 – Present
Leveraging structure in stochastic multi-armed bandit problems to optimize the use of human feedback. The human is modeled as a bounded-rational oracle capable of detecting arms deviating beyond a threshold from the optimal. The goal is to balance exploration and querying to efficiently eliminate suboptimal arms and minimize regret. Current focus: bandits with Lipschitz structure.

BanditsHuman FeedbackLipschitzRegret Minimization

Human-Guided Near-Optimal Regret Bounds Completed Nov 2024 – Mar 2025
Investigated whether side information from a bounded human oracle can improve theoretical guarantees of reinforcement learning. Focused on Upper Confidence Reinforcement Learning (UCRL): proposed an initial realistic user model and improved the regret guarantee of UCRL, providing a theoretical bound on the gain achievable through human assistance.

UCRLHuman OracleRegret BoundsTheory

Lowering Cognitive Burden in RLHF for Drug Design Completed Summer 2023 – Summer 2024
Proposed preferential feedback types to reduce cognitive burden on human annotators, leveraging the insight that humans find it easier to compare two instances than to evaluate a single one. Introduced a user model to effectively learn the reward using these new query types.

RLHFDrug DesignCognitive LoadReward Learning

Federated GTVMin Stochastic Multi-armed Bandit On Pause Aug 2025 – Oct 2025
Proposed a Best Arm Identification (BAI) approach utilizing neighbouring agent estimates in a decentralized network of regret-minimizing agents. Employs a global total-variation-based minimization strategy as a recommender, establishing a baseline for BAI with minimal regret in federated settings.

Federated LearningBanditsBAIDecentralized

Upper Confidence Reinforcement Learning exploiting state-action equivalence Completed 2019
Introduced a notion of similarity of state-action pairs based on the behavior of transition structures in order to reduce the regret bound of current approaches in reinforcement learning with single trajectory and unknown MDP. Proposed a method of clustering to group state-action pairs and aggregate group samples.

UCRLState-Action EquivalenceClusteringMDP

Online Personalized Mean Estimation in Collaboration Completed 2021 – 2022
In a network of agents each solving a personalized mean estimation problem, each agent receives a sample directly from its distribution and chooses a neighbour to communicate with. Using the Optimism in the Face of Uncertainty principle, proposed an algorithm to estimate the class of similar agents for each, with theoretical guarantees on class estimation and convergence time complexity.

Multi-agentMean EstimationOFUCollaboration

Distributed Multi-task Learning (M.Sc. Thesis) 2015 – 2016
We have solved the optimization problem of Multi-task learning in a distributed fashion while including the theoretical analysis of extending the framework on Policy Gradient Reinforcement Learning problems. We have reduced the convergence time of the algorithm and the memory consumption of each agent which makes this approach suitable for big data problems.

Multi-task LearningDistributedADMMPolicy Gradient

Online Face Recognition and Tracking 2016
We have developed an online face recognition software. After some preprocessing (illumination reduction and aligning the face) on each image of the video stream, we extract discriminative features and train a generative model so as to recognize the identity of passing individuals. Our approach is fast enough to be used for online processing.

Computer VisionFace RecognitionOnline Learning

Facial Expression Detection M.Sc. course project
Our goal was to guess human emotions using facial expressions. Employed image processing techniques to extract initial features and used epsilon-SVR to transfer them into a more discriminative space. A model for facial emotion detection was constructed by SVM that returns the probability of each emotion category.

Emotion RecognitionSVMComputer Vision

Predicting Excitement for DonorsChoose.org — KDD Cup 2014 — Top 25% 2014
Given projects' information from DonorsChoose.org, predicted how exciting each project is. Using data mining techniques, extracted relevant and discriminative features. Investigated sampling methods to overcome the imbalanced dataset problem and applied random forest, naive bayes, and other methods.

Data MiningClassificationImbalanced Data

Malware Detection using XCS M.Sc. course project
The goal was to distinguish malicious and benign binary files using: 1) XCS (RL-based Genetic Algorithm) and 2) N-gram based feature weighting using Genetic Algorithm.

SecurityGenetic AlgorithmXCS

Fuzzy Rule-Based Classification System M.Sc. course project
Learned the rule weights of a fuzzy rule-based classification system using a three-step greedy simulated annealing approach. This reduces the number of rules required while remaining fast and efficient.

Fuzzy LogicSimulated AnnealingClassification

Behavior-Selection by Emotions and Cognition in a Multi-Goal Robot Task B.Sc. Project
Designed a learning agent that finds a policy to live longer while encountering conflicting situations, implementing a reinforcement learning-based agent that maintains high performance while satisfying conflicting goals.

Reinforcement LearningRoboticsMulti-goal

• Reviewer:
  • 2025, 2026 - Reinforcement Learning Conference (RLC)
  • 2024 - Present - Transactions on Machine Learning Research (TMLR)
• Recruiting Committee Member:
  • 2023 - 2025 - Postdoc, PhD and Intern Hiring at Aalto

AAAI Bridge Program: Collaborative AI and Modeling of Humans

2023 – 2024 · Organizer

• 2020 Sep - Talk on Programming, Shiraz University

Hour of Code for elementary school students

2015 · Organizing Committee Member

• 2014 - Fall - Weekly presentations of AI
   We presented various topics: Natural Language Processing, Reinforcement Learning, Optimization, etc.

• 2009 - University Open Day program
  - Student chair of Computer Department

Women of Aalto

2026 – Present · President

• 2019 Jun - Talk on Academia Abroad, Shiraz University

TEDxShirazUniversity

2016 · Organizing Committee · Content Management & Media team

AIESEC

2016 · Team Member

• 2008 - 2012 - ACM student membership
• 2007 - 2011 - Shiraz University ACM association
• 2007 - 2009 - Computer Department's "Students' Scientific Council"

Feel free to reach out — I'm always happy to chat about research, collaborations, or ideas.