Andrews Cordolino Sobral, Ph.D.

Summary
Last page update: 11/08/2022
— Ph.D. on Computer Vision and Machine Learning.
— Head of AI at ActiveEon, Paris - France.
— Computer Vision Specialist (10+ years of experience).
— Experienced C++ and Python Developer.
— Proactive, problem-solver and self-learning skills.
— Contact: andrewssobral at gmail dot com
20+ years of professional experience with software engineering, software architecture, and software design. I have been working since Nov 2000 with low-level and high-level software development at several IT companies.
For more info, please see my CV.
Currently, I'm working as a Head of the Artificial Intelligence (AI) team at ActiveEon by focusing on the intersection of scientific research and software engineering, bringing cutting-edge AI technologies to the core platform of ActiveEon.
My main focus is on AI at Scale, HPC+IA, and MLOps.
From 2017, my team and I worked on the development of the Proactive AI Orchestration platform, helping customers to automate and orchestrate AI-based workflows, scaling-up with parallel and distributed execution.
My responsibilities include defining the AI roadmap, supporting AI team members, and driving AI key projects.
In short, my activities are:
— Lead a team of 4 (four) PhDs on AI & Machine Learning;
— Drive AI roadmap creation and execution;
— Drive end-to-end AI product design, architecture, and operation;
— Develop end-to-end AI workflows;
— Develop large-scale AI workflows with parallel & distributed execution;
— Develop multi-node and multi-GPU AI workflows;
— Develop a distributed AutoML tool for large-scale hyperparameter optimization and neural architecture search;
— Develop the Proactive Python SDK to submit runtime dynamic AI workflows;
— Develop the Proactive Jupyter Kernel to submit runtime dynamic AI workflows from Jupyter Lab;
Summary of my AI research experience:
Since 2010, my AI research activities include intelligent video analysis for object detection, segmentation, and tracking, to perform event/action recognition and behavior classification/prediction.
I also developed and deployed computer vision and machine learning models at the edge on system-on-a-chip (SoC) devices such as Raspberry PI, PandaBoard, and NVIDIA Jetson boards (Nano, Xavier NX, TX2, and AGX Xavier).
My doctoral research was mainly focused on the R&D of advanced matrix and tensor methods for robust low-rank/sparse representation and subspace learning of multidimensional and streaming data, such as moving object detection and background-foreground separation in multi-spectral and multi-featured video sequences.
In the past, between my B.Sc. and my M.Sc, I worked on several embedded systems and robotic projects using microcontrollers and SoC boards to read sensors, I/O devices, performing data acquisition, signal processing, image/video processing, wireless communication, and motor control.
For more information, please see the Projects page at the top of this page.
Main research interests: Computer Vision, Machine / Deep Learning, Pattern Recognition, Applied Mathematics, Matrix / Tensor Decomposition, Optimization, Big Data.
Journal reviewer: I peer-reviewed for more than 10 high-quality journals, such as Elsevier (CVIU, IVC, PRL, NC, IF), Springer (NCA, CC, FITEE, JIVP), IEEE (TIP, TNNLS, TCSVT), MDPI (Sensors), JEI, JOSA-A, PLOS ONE. For more information, please see my Publons profile for an updated version.
Int. Conferences/Workshop reviewer: ICCV 2015 (RSL-CV), ICIAP 2015 (SBMI), ISCV 2017.
My open source libraries and source code:
BGSLibrary - A Background Subtraction Library.
VDTC - Vehicle Detection, Tracking, and Counting.
LRSLibrary - Low-Rank and Sparse tools for Background Modeling and Subtraction in Videos.
MTT - Matlab Tensor Tools for Computer Vision.
IMTSL - Incremental and Multi-feature Tensor Subspace Learning.
OSTD - Online Stochastic Tensor Decomposition.
For more details, please visit my GitHub profile: https://github.com/andrewssobral






