cv

Visiting Researcher

ServiceNow

Fine-tuning large language models using Discrete Diffusion models.

• Developed and implemented innovative models in diffusion

Software Engineer

Cafe Bazaar Inc. - Divar

Full stack developer for the chat team, responsible for online chat and call services.

• Developed AI models for detecting harassment and scams.

Machine Learning Engineer

Yektanet Inc.

Started as a data scientist in the Data team and was quickly promoted to a machine learning engineer for the Infrastructure team.

• (@ Data team) Built content-based recommenders for ads and personalization.
• (@ Data team) Designed a geolocation system, increasing CTR by 6.1%.
• (@ Infrastructure team) Developed a real-time data pipeline for ad serving, improving latency by 30%.
• (@ Infrastructure team) Implemented a distributed system for large-scale data processing, enhancing scalability.
• (@ Infrastructure team) Owned and maintained backend and database services for internal tooling.

M.Sc.

MILA - Quebec AI Institute and McGill University

Electrical and Computer Engineering

• Multi-agent systems for counterfactual generation and visual explainability. (submitted work @ iccv workshops
• Diffusion and GAN-based approaches for counterfactual image synthesis. (ORAL@ MIDL`24 and short-listed for Best Paper Award)
• Bias mitigation for improving robustness and generalization in medical imaging. (Published work @ MICAAI)
• Longitudinal generative modeling and trajectory forecasting for irregularly sampled time series.

B.Sc.

Sharif University of Technology

Computer Engineering

• Co-authored one of the earliest works applying Vision Transformers to medical imaging, establishing a benchmark for subsequent research.

Unifying Autoregressive And Diffusion-Based Sequence Generation

ICLR’25 - DelTa

DeCoDEx: Confounder Detector Guidance for Improved Diffusion-based Counterfactual Explanations

MIDL’24 ORAL

Shortlisted for Best Paper Award [Top 3.6%]

Debiasing Counterfactuals In the Presence of Spurious Correlations

FAIMI - MICCAI’23

SwinCheX: Multi-label classification on chest X-ray images with transformers

arXiv

Towards human pose prediction using the encoder-decoder lstm

ICCV Workshops’21