About

  • City: Vancouver, Canada
  • Degree: Master's in Computing Science

I've always been passionate about stories, especially visual ones. My first love was watching movies and series, and I could spend hours engrossed in them. The power of these visual stories to captivate and convey complex ideas has always fascinated me.
From an early age, I discovered a passion for math and logical thinking. I found coding to be a natural extension of these interests, presenting both logical and creative challenges that I found incredibly engaging.
In my undergrad, I found a way to merge these two worlds: Computer Vision combined with Machine Learning. This realization opened up a whole new realm of possibilities where I could blend my love for visual storytelling with my technical skills.
To this day, the field of computer vision continues to captivate me. Whether it’s enhancing images, creating new visual content, or developing innovative solutions, the intersection of storytelling, math, and coding through computer vision is endlessly inspiring to me.
When I'm not coding, I love being in nature, especially by the ocean, and staying active. I also enjoy exploring ideas and new places.

Interests

Programming

Movies

Artifiacl Intelligence

Paddle Boarding

Improv

Traveling

Hiking

Board Games

Resume

Summary

Sepi (Sepideh) Sarajian Maralan

  • Vancouver, BC, Canada
  • sepide[dot]sarajian[at]gmail[at]com

Education

Master of Computing Science

2020 - 2022

Simon Fraser University, BC, Canada

Thesis: Computational Flash Photography
Related Courses:
Machine Learning: A+, Computer Vision: A+, Design and Analysis of Algorithms: A, Computational Photography: A, Geometric Modelling in Computer Graphics: A

Bachelor of Computer Engineering

2014 - 2019

Sharif University of Technology, Tehran, Iran

Related courses:
Computer Graphics, Advanced 3D Computer Vision, Digital Signal Processing, Design of Algorithms, Advanced Programming, Data Structures

Professional Experience

Senior Machine Learning Engineer

2025 - Present

Novarc Technologies

  • Real-Time Vision System: Developed state-of-the-art real-time line detection for robotic applications.
  • Model Optimization & Deployment: Optimized deep learning models for deployment on embedded systems, ensuring low latency and high reliability in production environments.

Intern & Machine Learning Research Engineer

2022 - 2025

Lipdub AI

  • Advanced Speech-to-Lip Synchronization: Develop state-of-the-art lip-sync technology, employing cutting-edge network architectures such as Vision Transformers and Diffusion models. Focus on enhancing the accuracy and naturalness of speech-to-lip generation for dynamic content.
  • Quality Assurance and Bug Resolution: Identify and resolve potential bugs and edge cases. Ensure the robustness and reliability of the lip-sync model, delivering seamless and error-free results.
  • Cost Reduction in AI Processing: Reduced AI processing costs by 25% through optimization of data pipelines, computational resources, and model architectures, enabling faster and more cost-effective development cycles.

Research Assistant

2020 - 2022

GrUVi lab, Simon Fraser University

  • Implemented a novel architecture for Image Relighting using Intrinsic Decomposition and deep neural networks, generating high-resolution and high-quality output.
  • Conducted image preprocessing and compositing of diverse datasets, proficiently harmonizing, and combining portrait images for seamless integration.

Projects

Computational Flash Photography

Computational Flash Photography Through Intrinsics

A system to computationally control the flash light in photographs originally taken with or without flash.

Accepted at CVPR 2023.

Dolly Zoom 1
Dolly Zoom 2

Dolly Zoom

Dolly Zoom effect created on a single RGB image with monocular depth estimation.

2021

Mix and Match

Mix and Match

An automated mix-and-match part assembly-based 3D modeling.

2021