Research interests
- LLM agents, computer-use agents, and agent evaluation
- Self-improving and memory-augmented web agents
- Vision-language models, open-vocabulary learning, and semantic segmentation
- Vision transformer pre-training
Education
- Ph.D. in Artificial Intelligence | ETS Montreal | 2022 - 2026
- M.Sc. in Artificial Intelligence | Sharif University of Technology | 2018 - 2021
- Supervisor: Professor Mahdieh Soleymani Baghshah
- Dissertation Title: Representation Learning using Deep Neural Networks and Information Theory
- B.Sc. in Software Engineering | Sharif University of Technology | 2014 - 2018
Work experience
- Visiting Researcher
- ServiceNow AI Research: https://www.servicenow.com/research
- September 2025 - June 2026
- Developed and evaluated self-improving web agents that learn reusable skills, workflows, and memories from past trajectories without fine-tuning.
- Studied web-agent memory and skill modules under controlled inference-token budgets in WebArena/BrowserGym-style environments.
- Contributed to AgentHorizon, a benchmark for evaluating agentic judges on long-horizon computer-use workflows.
- Researcher
- LIVIA, École de technologie supérieure, Montreal
- January 2022 - present
- Published first-author work on few-shot segmentation, open-vocabulary segmentation, and locality-aware ViT pre-training.
- Built and released PyTorch research code for segmentation, CLIP/VLM adaptation, and ViT pre-training projects.
- Software Engineer
- Divar Classified Ads: https://divar.ir
- July 2016 - October 2018
- Built and maintained production microservices and web applications using Python, Django, Docker, Redis, PostgreSQL, and Golang.
