Benoit Hamelin Email & Phone Number

Ottawa, Ontario, Canada

Machine learning applied to cyber defense. Bad guys, put on your best black hat, because I see you.

Montreal, Canada Area

I lead the technical development of cybersecurity products designed AI-first.

I lead the development of a large-scale distributed computing system for detecting, intercepting and responding to advanced persistent threats.

Collaborative development of a large-scale data analysis infrastructure for defending networks from targeted intrusions known as "advanced persistent threats" (APT). This involved the development of lightweight sensor software deployed on our clients' machines, a multi-platform project of some 20000 lines of C code. In addition, I contributed to the development of a distributed data storage and analysis back-end written in Python, leveraging cutting-edge technologies such as the Cassandra… Show more Collaborative development of a large-scale data analysis infrastructure for defending networks from targeted intrusions known as "advanced persistent threats" (APT). This involved the development of lightweight sensor software deployed on our clients' machines, a multi-platform project of some 20000 lines of C code. In addition, I contributed to the development of a distributed data storage and analysis back-end written in Python, leveraging cutting-edge technologies such as the Cassandra database system. I coordinated the work of one senior exploit researcher as well as up to six junior programmers. Show less

I researched 3D tissue imaging by diffuse optical tomography of human soft tissue using 2D illumination and detection patterns. My work consisted in finding sets of illumination patterns that enhanced the quality of reconstructed images. The methods I developed (yet to be published) provided pattern sets that yielded better image resolution close to the boundaries of the tissue under examination, as well as deeper sensitivity.

Montréal, Québec, Canada

Research: fast reconstruction in X-ray tomography with metal artifact reductionClinical methods for restenosis detection in post-angioplasty patients are either of low accuracy (2D angiography) or invasive (intra-veinous ultrasonography). 3D angiography through computed tomography overcomes both issues, but intra-arterial metal stents grafted during angiography complicate CT imaging by introducing metal artifacts that prevent accurate measurement of the artery's lumen. Previous work… Show more Research: fast reconstruction in X-ray tomography with metal artifact reductionClinical methods for restenosis detection in post-angioplasty patients are either of low accuracy (2D angiography) or invasive (intra-veinous ultrasonography). 3D angiography through computed tomography overcomes both issues, but intra-arterial metal stents grafted during angiography complicate CT imaging by introducing metal artifacts that prevent accurate measurement of the artery's lumen. Previous work proposed a CT reconstruction method that eliminated metal artifacts, but took hours of computation. Through my research, I achieved a reduction of the runtime by a full order of magnitude (from approximately 5 hours to about 20 minutes) by choosing an alternative statistical data formation model, by developing targeted reconstruction around the artery of interest and by running the computation using a distinct, faster numerical method.I taught signal and image processing courses at both the undergraduate and graduate levels. I succeeded at making these highly mathematical courses accessible and enjoyable even to medical engineering courses, who often enlisted with basic or obsolete knowledge in mathematics. I also wrote supplemental material to better support the students, who rated it positively. Show less

Sherbrooke, Québec, Canada

Research: applications of automatic differentiation to nonlinear optimizationNumerical methods for nonlinear optimization rely on the finite difference between the evaluation of the objective function at the current and at the next point in the trial sequence. A naive computation of this difference may yield numerical inaccuracy that prevent proper detection of convergence. My research derived an approach from the technique of automatic differentiation to compute this difference without… Show more Research: applications of automatic differentiation to nonlinear optimizationNumerical methods for nonlinear optimization rely on the finite difference between the evaluation of the objective function at the current and at the next point in the trial sequence. A naive computation of this difference may yield numerical inaccuracy that prevent proper detection of convergence. My research derived an approach from the technique of automatic differentiation to compute this difference without any numerical issue. Thus modified, off-the-shelf numerical methods often exhibited better numerical behavior, converging faster.Teaching of math and software development classes at the undergraduate level. I wrote all material for an optional course to avoid students the purchase of expensive textbooks. I also put all my course material (notes, homework, exam solutions) on the web, before campus-wide academic CMSs made this easy. Students commended my dedication and the quality of my teaching. I was also proud to help struggling students back on their feet, so they'd perform at least one full grade level higher than they did before we met. Show less

Montreal, Québec, Canada

My associates and I worked on a time tracking program for accountants, on Windows. Our software involved an easy-to-use Excel-like grid as main UI, and provided a template-based report generator. Although the application was close to achieving its goals, financial issues forced us to shelve the project before we could release an initial version.

Design and program software for surgery assistance and support.

Design and program software for casual illustration and imaging product.

Doctor Of Philosophy (Ph.D.), Medical Imaging

Graduated with skills in research, signal and image processing, medical imaging and numerical optimization.

Master Of Science (M.Sc.), Computer Science

Graduated with Honors. Skills in research, numerical methods and nonlinear optimization.

Bachelor Of Science (B.Sc.), Computer Science

Activities and Societies: AGES, Impérius Graduated with top marks in Computer Science, with fundamental computation and software.

Education record

Education record