Resume

Experience

Computer Vision Engineer — Quividi

January 2024 — July 2024 · Paris, France

Worked in R&D on CNN-based attribute-classification models for edge-oriented deployment.

• Evaluated classification architectures against deployed baselines with different speed and accuracy trade-offs
• Developed automatic data-labelling tooling and a training/fine-tuning framework
• Recommended architectures appropriate for edge devices
• Integrated the selected model into the main software product

Environment: Python, PyTorch, OpenCV, dlib, C++, OpenVINO, TensorBoard, Linux, Git

Image Processing Engineer — Essilor

March 2023 — January 2024 · Créteil, France

Contributed to software for visual-system diagnosis, imaging, and measurement.

• Developed, evaluated, and integrated algorithms for corneal-topography reconstruction, biometry, and aberrometry
• Compared nonlinear optimisation methods for reconstruction workflows
• Integrated eye-depth measurement based on optical coherence tomography
• Improved biometry processing through short-time Fourier-transform-based artifact removal

Environment: C++, Python, OpenCV, Eigen, NLOpt, NumPy, SciPy, Qt, Jenkins, GitHub, Jira

Data Scientist — L’Oréal

January 2023 — February 2023 · Remote

Developed statistical tooling for clinical-study analysis.

• Interpreted the Statistical Study Plan and translated requirements into analysis tools
• Built tools for descriptive analysis, normality checks, and frequentist inference

Environment: R, RStudio

Computer Vision Engineer — L’Oréal

May 2022 — December 2022 · Clichy, France

Contributed to an augmented-reality system that simulated hair colour using camera sensing and projector illumination.

• Integrated camera-projector calibration, hair detection and segmentation, UI components, and the colouring model
• Developed automatic hair detection and segmentation using depth and colour information
• Improved calibration guidance, application behaviour, and user experience through debugging and integration work

Environment: C++, Python, OpenCV, Intel RealSense, Geogram, ImGui, GLFW, Git

Consultant — Meritis

May 2022 — January 2024 · Paris, France

Delivered assignments for L’Oréal and Essilor across computer vision, image processing, and statistical analysis.

• Developed signal- and image-processing algorithms for Essilor
• Built statistical tools supporting L’Oréal clinical studies
• Contributed to an augmented-reality dynamic projection system for L’Oréal

Early-Stage Researcher — L2S, CentraleSupélec, CNRS, Université Paris-Saclay

December 2018 — February 2022 · Saclay, France

Conducted doctoral research within the European RealVision Marie Skłodowska-Curie project on learning-based coding for light-field imaging.

• Developed hybrid learning-based and traditional video-coding approaches for lossy light-field compression
• Proposed a learning-based method for lossless light-field compression
• Investigated distributed source coding for light fields
• Published a doctoral dissertation, book chapter, conference papers, and journal article
• Collaborated with international research partners across the RealVision network

Environment: Python, TensorFlow, PyTorch, OpenCV, scikit-learn, NumPy, TensorBoard, MATLAB, Linux, Git, LaTeX

R&D Engineer — Protech

February 2015 — November 2018 · Novi Sad, Serbia

Developed computer-vision and image-processing solutions for external customers, including real-time and resource-constrained systems.

• Built text-segmentation processing that generated precise binary maps for OCR
• Developed basketball-player detection and recognition systems, including annotation tooling, feature-based models, and later deep-learning approaches
• Optimised CV workloads for ARM and x86 hardware, including model pruning and lower-precision inference
• Evaluated and deployed pedestrian-tracking methods
• Built document-recognition systems using keypoints and descriptors for mobile-device constraints
• Developed and deployed an SVM-based legume-quality estimation system that met accuracy and real-time requirements
• Presented technical work to customers and supported junior team members

Environment: C++, Python, OpenCV, VLFeat, Keras, Caffe, MATLAB, Mercurial, Jira, SIMD, ARM NEON

Image Processing Intern — Agfa-Gevaert

August 2013 — August 2014 · Antwerp, Belgium

Explored a generic image-enhancement system for newspaper-printing workflows.

• Evaluated a commercial image-enhancement product and identified limitations in its processing workflow
• Proposed an updated architecture and processing operations to increase autonomy
• Developed approaches for identifying artificial images and normalising inputs for subsequent enhancement
• Improved processing quality for images reported by customers as unacceptably processed

Environment: C++, MATLAB, OpenCV, Eigen

Education

Doctor of Philosophy in Signal and Image Processing

2018 — 2022 · Université Paris-Saclay and CentraleSupélec

Doctoral research on learning-based coding for light-field imaging.

Master of Science in Electrical and Electronic Engineering

2012 — 2014 · EPFL, Swiss Federal Institute of Technology Lausanne

Bachelor of Science in Electrical Engineering

2008 — 2012 · University of Novi Sad, Faculty of Technical Sciences

Selected publications

• Stepanov, M. (2022). Selected topics in learning-based coding for light-field imaging. Doctoral dissertation, Université Paris-Saclay.
• Stepanov, M., Valenzise, G., & Dufaux, F. (2022). Compression of light fields. In Immersive Video Technologies.
• Stepanov, M., Valenzise, G., & Dufaux, F. (2020). Hybrid learning-based and HEVC-based coding of light fields. IEEE ICIP.
• Stepanov, M., Mukati, M. U., Valenzise, G., Forchhammer, S., & Dufaux, F. (2021). Learning-based lossless light field compression. IEEE MMSP.
• Mukati, M. U., Stepanov, M., Valenzise, G., Dufaux, F., & Forchhammer, S. (2020). View synthesis-based distributed light field compression. IEEE ICME Workshops.
• Mukati, M. U., Stepanov, M., Valenzise, G., Forchhammer, S., & Dufaux, F. (2021). Improved deep distributed light field coding. IEEE Open Journal of Circuits and Systems.

Technical skills

Core areas

Computer vision, image processing, machine learning, deep learning, image compression, optical and medical imaging, statistical analysis, real-time and edge deployment.

Languages and tools

Python, C++, C, MATLAB, R, PyTorch, TensorFlow, OpenCV, scikit-learn, NumPy, SciPy, OpenVINO, Docker, Linux, Git, GitHub, Jenkins, Jira.

Languages

• Serbian — Native
• English — Professional working proficiency
• French — Professional spoken proficiency; intermediate written proficiency