cv

Freelance Machine Learning Developer

Rocket AB

Remote freelance work based in Stockholm, Sweden, developing multi-modal searching pipeline and backend infrastructure.

• Developed Multi-modal (image + text) Searching pipeline using CLIP embeddings for online/low-latency item retrieval from NoSQL database of >500,000 items
• Developed backend API using FastAPI achieving few millisecond latencies for retrieval
• Dockerized and deployed final pipeline on Azure

Lead ML Researcher

Hospital for Special Surgery (HSS)

Leading ML in musculoskeletal imaging research projects in collaboration with Dr. Ayoosh Pareek's Machine Learning team. Co-secured over $70,000 in competitive research funding across three major grants.

• Developed state-of-the-art deep learning models for bone age prediction from hand X-rays, achieving ~3-month error (up to 20% improvement over previous SOTA)
• Pioneered novel approach for bone age estimation from knee X-rays, achieving ~5-month error
• Developed deep learning model for automated tibial spine landmark detection on knee radiographs achieving <1% keypoint detection error
• Published findings in American Journal of Sports Medicine (AJSM) and presented at ISAKOS 2025

Machine Learning Developer

Aris Co.

Developed ML-powered application for skin health product recommendation in Tehran, Iran.

• Developed robust deep learning models leading to 2x improvement in metrics (face detection and segmentation, landmark detection, skin condition classification)
• Developed core ML-powered application for skin health product recommendation
• Collected and annotated multiple image datasets
• Developed backend API using Django REST framework achieving <20ms latency on API calls at scale

Workshop Lecturer

SIIM Annual Meeting

Workshop on 'Towards Trustworthy Automated LLMs' at SIIM Annual Meeting, Baltimore, U.S.

Invited Reviewer

ICML/ICLR Workshops + PLoS ONE

Peer review service for top machine learning conferences and medical journals.

M.D. (Medical Doctor)

Tehran University of Medical Sciences (TUMS)

Medicine

ISHA Research Grant

ISHA

$20,000 grant for automating morphology parameters on hip imaging

ON Education Grant

ON Foundation

$28,000 grant for osteoarthritis progression prediction using diffusion models

AOSSM Playmaker Grant

AOSSM

$25,000 grant for AI-based bone age prediction from hand & knee radiographs

Selected Participant, M2L Summer School

Google DeepMind

Sponsored participation in M2L Summer School, Split, Croatia

Travel Grant + Registration Waiver, EEML Summer School

Google DeepMind

EEML Summer School, Novi Sad, Serbia

Gold Grant, USERN PID Summer School

USERN

Zurich, Switzerland

Shortlisted for Trainee Research Prize

Radiological Society of North America (RSNA)

Recognition for research excellence in radiology

1st Place, Data Science Competition

Digikala

First place in data science competition

3rd Place, Code Cup

Quera

Third place in coding competition

Solo Bronze Medal, Shopee Competition

Kaggle

Bronze medal in Shopee competition

Solo Bronze Medal, Basic Sciences Olympiad

Iran's Ministry of Health

Bronze medal in national olympiad

Fully Funded MD Scholarship

National University Entrance Exam

Ranked 39th out of >600,000 participants (top 0.01%)

Deep learning Models, How to Develop and Deploy (Book Chapter)

Basic Methods Handbook for Clinical Orthopaedic Research

Book chapter on developing and deploying deep learning models in clinical orthopaedic research. Forthcoming 2026.

Determination of Skeletal Age From Hand Radiographs Using Deep Learning

American Journal of Sports Medicine (AJSM)

Deep learning approach for skeletal age determination from hand radiographs with state-of-the-art accuracy.

Towards trustworthy artificial intelligence in musculoskeletal medicine: A narrative review on uncertainty quantification

Knee Surgery, Sports Traumatology, Arthroscopy (KSSTA)

Narrative review on uncertainty quantification methods for trustworthy AI in musculoskeletal medicine.

LatteReview: A Multi-Agent Framework for systematic review automation using large language models

arXiv.org (preprint)

Multi-agent LLM framework for automating systematic literature reviews.

CONFLARE: CONFormal LArge language model REtrieval

arXiv.org (preprint)

Conformal prediction approach for large language model retrieval systems.

Predictive modeling for acute kidney injury after percutaneous coronary intervention in patients with acute coronary syndrome: a machine learning approach

European Journal of Medical Research

Machine learning model for predicting acute kidney injury following percutaneous coronary intervention.

The role of pre-training data in transfer learning

Workshop on Multimodal Representation Learning, ICLR 2023

Investigation of how pre-training data distribution affects transfer learning performance in vision models.

TransDeepLab: Convolution-Free Transformer-Based DeepLab v3+ for Medical Image Segmentation

PRIME workshop, MICCAI 2022

Novel transformer-based architecture for medical image segmentation achieving state-of-the-art performance. Accumulated ~200 citations.

Deep learning model for measurement of shoulder critical angle and acromion index on shoulder radiographs

JSES Reviews, Reports, and Techniques

Automated deep learning approach for measuring shoulder anatomical parameters from radiographs.

How well do contrastively trained models transfer?

Workshop on Pre-training: Perspectives, Pitfalls, Paths Forward at ICML 2022

Analysis of transfer learning performance in contrastively trained vision models with novel efficiency improvements.

Deep Learning for Musculoskeletal (MSK) Imaging

• Co-secured over $70,000 in competitive research funding
• Achieved state-of-the-art bone age prediction with ~3-month error from hand X-rays
• Pioneered bone age estimation from knee X-rays with ~5-month error
• Published in American Journal of Sports Medicine

Transfer Learning in Self-Supervised Vision Models

• Designed and executed large-scale experiments on cluster of 8 V100 GPUs
• Invented novel method to predict downstream transfer performance with >10x computational efficiency
• Co-authored papers at ICML 2022 and ICLR 2023 workshops

Transformer-Based Medical Image Segmentation

• Re-engineered DeepLab v3+ with Swin-Transformer blocks
• Achieved state-of-the-art performance across multiple medical imaging datasets
• Paper accumulated ~200 citations

PyTorch CLIP Tutorial

• 700+ GitHub stars
• Widely used educational resource