Computer Vision Researcher · M.Sc.

I work on computer vision and perception, particularly on problems that sit between research and reality.

I enjoy the part where things stop being clean: when a method that works on a benchmark has to survive different sensors, imperfect data, limited compute, unfamiliar hardware, and the constraints of a real product. Much of my work has been in automotive and in-cabin perception, but I'm interested in perception more broadly.

Over the years, I've moved between research, applied engineering, project leadership, and technical leadership. I like difficult problems, questioning assumptions, and building things that are useful beyond a paper or a demo.

Seoul, South Korea · Originally from Bangladesh

My work has largely centered around understanding the world through visual and sensor data.

• Driver and in-cabin monitoring
• Detection and segmentation
• Monocular and stereo depth
• 3D reconstruction and geometric perception
• Body keypoints and human pose
• Multi-sensor perception

• Model compression and optimization
• Edge and embedded deployment
• Platform porting
• Production pipeline integration
• Real-time perception systems

• Computer vision
• Generative vision
• 3D vision
• Event-based vision
• Multi-modal and multi-sensor learning

Primary stack: Python · C++ · PyTorch · OpenCV · ONNX · TFLite · ROS

Some of the problems I've worked on across research and industry.

In-cabin depth estimation Monocular depth estimation for production in-cabin monitoring systems. Deep In Sight · SL / Mobase / Mobis · 2024–2026

3D human understanding Body keypoints, seat pose, gaze, and child-presence estimation for in-cabin units. Deep In Sight · SL / Mobase · 2024–2026

Level 4 ADAS perception Pedestrian detection, distance estimation, and predictive trajectory estimation for an autonomous-driving PoC. DeltaX · KADIF · 2023–2027

3D display perception Real-time eye-distance estimation from a 3D light-field display. DeltaX · Hyundai Mopic · 2023

Long-wave infrared perception Worker detection, localization, and tracking using thermal imagery. DeltaX · Korea Railroad Research Institute · 2022–2023

Event-based stereo Deep stereo matching and motion deblurring using stereo event cameras. CVIP Lab · SK hynix · 2021–2022

Holographic display perception High-speed pupil tracking for a holographic display. CVIP Lab · ETRI · 2019–2020

Before moving deeper into applied automotive perception, much of my work focused on 3D vision, image restoration, event-based vision, and generative models.

Multi-Scale Attention-Guided Non-Local Network for HDR Image Reconstruction Sensors · 2022

Unsupervised Deep Event Stereo for Depth Estimation IEEE Transactions on Circuits and Systems for Video Technology · 2022

SIFNet: Free-Form Image Inpainting Using a Color Split-Inpaint-Fuse Approach Computer Vision and Image Understanding · 2022

Deep Event Stereo Leveraged by Event-to-Image Translation AAAI-21 · 2021

Global and Local Attention-Based Free-Form Image Inpainting Sensors · 2020

I also enjoy the competitive side of research — taking a problem with a fixed evaluation protocol and trying to push the limits.

I led the DITM entry for the AIM 2025 inverse tone-mapping challenge. Code →

Senior AI/ML Researcher · R&D Nov 2024 – Present

Working on production perception systems for driver and in-cabin monitoring, spanning research, model development, optimization, deployment, and system integration.

AI Researcher → Project Lead → Group Lead · Automotive Perception Oct 2022 – Nov 2024

Worked across automotive perception while progressively taking on technical and project leadership. Projects covered driver and occupancy monitoring, depth and 3D perception, sensor fusion, and downstream perception systems.

Vision Researcher Mar 2019 – Sep 2022

Research across depth estimation, 3D reconstruction, event-based vision, image restoration, and generative vision.

Business Development Analyst Jul 2018 – Jan 2019

Teaching Assistant Feb 2018 – Jun 2018

M.Eng. · IT Convergence Engineering Gachon University · 2019–2021

Thesis: deep-learning-based image inpainting for irregular masks using attention.

B.Eng. · Electronics and Telecommunication Engineering University of Liberal Arts Bangladesh · 2013–2017

I'm interested in opportunities where technical depth actually matters.

• Early-stage companies building serious technical products
• Founding roles where I can work across research, product, and engineering
• Technical leadership at the boundary of research and implementation
• Research collaborations on difficult perception problems
• Open-source work that is genuinely useful

I'm less interested in doing research for the sake of producing another benchmark result, and more interested in finding problems where the research has somewhere to go.

Languages Python · C++

Deep learning & vision PyTorch · TensorFlow · Keras · ONNX · TFLite · OpenCV · Open3D

Systems & robotics ROS · Docker · Git

Other NumPy · SciPy · scikit-learn · Pandas · Jira · Confluence · Azure DevOps · Notion

Email smnadimuddin [at] gmail [dot] com

Elsewhere Website · LinkedIn · Google Scholar · ResearchGate