cv
General Information
| Full Name | Alireza Moazeni |
| Website | amoazeni75.github.io |
| Research Interests | Computer Vision & 3D Neural Rendering, Generative Modeling & Generative AI, Self-Supervised & Unsupervised Representation Learning, Large Language Models & Natural Language Interfaces |
Education
-
Sep 2021 - Present Ph.D. in Computer Science
Simon Fraser University (SFU), British Columbia, Canada - GPA: 4.17/4.33
- Supervisor: Prof. Ke Li
-
Sep 2015 - Sep 2020 Bachelor of Science in Computer Engineering & Electrical Engineering
Amirkabir University of Technology (AUT), Tehran, Iran - Major: Computer Engineering
- Second Focus: Electrical Engineering (Control)
- GPA: 18.40/20 (3.84/4.00)
- Ranked in top 5% among ~100 undergraduate students in Computer Engineering and IT Department
- Ranked 4th among ~35 undergraduate students in Electrical Engineering (Control)
Publications
-
Preprint PAPR Cut: Runtime Optimizations for Efficient Point-Based Neural Rendering
- Authors: Alireza Moazeni, George Shramko, Shichong Peng, Yanshu Zhang, Ke Li
-
Submitted to CVPR 2026 Intrinsic PAPR for Point-level 3D Scene Albedo and Shading Editing
- Authors: Alireza Moazeni, Shichong Peng, Yanshu Zhang, Chiraq Vashist, Ke Li
- Submitted to CVPR 2026
-
Submitted to ICLR 2026 PROSE: Point Rendering of Sparse-Controlled Edits to Static Scenes
- Authors: Yanshu Zhang, Shichong Peng, Mehran Aghabozorgi, Alireza Moazeni, Ke Li
- Submitted to CVPR 2026
-
Submitted to CVPR 2026 Illuminating Uncertainty in Masked Image Models
- Authors: Tristan Engst, Alireza Moazeni, Ke Li
- Submitted to CVPR 2026
-
2023 PAPR: Proximity Attention Point Rendering
- Authors: Yanshu Zhang*, Shichong Peng*, Alireza Moazeni, Ke Li
- Neural Information Processing Systems (NeurIPS) - Spotlight paper
-
2022 CHIMLE: Conditional Hierarchical IMLE
- Authors: Shichong Peng, Alireza Moazeni, Ke Li
- Neural Information Processing Systems (NeurIPS)
Industrial Experience
-
Oct 2021 - Present Machine Learning Specialist and Team Leader (Part-time)
Canada - Lead ML specialist and team lead on multiple applied ML projects spanning embedded, medical, and cloud platforms
- Designed and deployed offline voice assistant models on embedded boards and smartphones, enabling on-device speech recognition and intent understanding without internet connectivity
- Led development of transformer-based models for blood pressure estimation and vital-sign forecasting from PPG, ECG, and other vital signals
- Co-designed and delivered the IoT2Cloud platform for web, iOS, and Android, integrating LLM-based natural language to UI-component generation with a custom evaluation framework and curated UI component dataset
- Optimized and deployed ML and LLM models under tight computational and memory constraints, improving reliability and latency for production use
-
May 2016 - Aug 2020 A.I. Developer and Product Manager
i-Cliqq Company, Shanghai, UAE. - Developed and launched Maya i-Box, an intelligent embroidery design system that converts photographs and touchscreen sketches into optimized DST stitch files for industrial embroidery and sequins machines
- Designed the computer vision and image-processing pipeline in C++/Qt on NanoPC embedded hardware, including color and noise reduction, edge and object detection, and shape analysis to robustly vectorize raster artwork
- Led algorithmic work in computational geometry and graph-based optimization to decompose complex shapes into stitchable polygons, improving automation quality, reliability, and performance for production workflows
Research Projects
-
Image2Blocks - Reconstructing images in 3D using cuboids
- Formulate monocular image understanding as reconstructing scenes using 3D cuboid primitives rather than only 2D attributes
- Combine monocular depth and point-cloud reconstruction to infer layout, free space, and occlusions in globally consistent 3D scene models
-
Text-Guided Diverse Image-To-Image Translation
- Develop model for text-guided image-to-image translation that preserves main object while applying text-specified transformations
- Leverage pre-trained language embeddings and conditional image synthesis to produce realistic, diverse outputs aligned with text
Skills
-
Programming & Systems
- Python, C++17, CUDA, Triton
- Linux, Git/GitHub, Docker, AWS
- GPU profiling and optimization
- Multi-GPU/distributed training (DDP, DeepSpeed)
- Mixed-precision training/inference
-
Deep Learning & Generative Modeling
- PyTorch, JAX
- Diffusion models, transformers, generative modeling
- Representation learning (MAE, contrastive/self-supervised)
-
LLMs & Efficient Inference
- LLM fine-tuning and distillation (LoRA/QLoRA)
- Pruning and compression
- Latency-optimized inference pipelines
- Prompt engineering and natural language interfaces
-
Computer Vision & 3D/Rendering
- Neural rendering
- Geometry processing
- COLMAP, Blender
- 3D intrinsic decomposition and inverse rendering
-
Tools & Experimentation
- WandB, TensorBoard, Slurm
- Reliable ML/LLM deployment in constrained environments
Honors and Awards
-
2025 - Lab2Market National Commercialization Program (Validate Stream), Canada
-
2023 - Helmut and Hugo Eppich Family Graduate Scholarship (Ebco Eppich Award Competition), Simon Fraser University, Canada
-
2020 - Top 5% among 100 undergraduate students in Computer Engineering and IT Department, Amirkabir University of Technology, Iran
-
2017 - Ranked 4th among 35 undergraduate students in Electrical Engineering (Control), Amirkabir University of Technology, Iran
-
2015 - Top 0.15% in Nationwide University Entrance Exam for B.Sc. in Math and Engineering (~250,000 applicants), Iran
-
2013 - 2nd place in RoboCup IranOpen 2013 - League Junior Soccer B Light Weight
- 1st place in RoboCup Helli (NODET) 2013 - League Junior Soccer B Light Weight
-
2010 - 1st place in National Conference on Mathematics, Analytic Geometry and Spatial Orientation, Ministry of Science, Iran