Jasmeet Singh

⋆ Worked with the web applications backend team on critical client facing applications named - Patient Signals and Patient Image Search
Directed a cross-functional team in the development and launch of an advanced image neighborhood search product, enhancing functionality and user experience.
⋆ Spearheaded innovations that reduced patient cohort identification time for pharmaceutical clients from several days to mere minutes.
⋆ Implemented A/B testing to optimize features, including plug-and-play AI models for image similarity and interactive similarity graph visualizations.
⋆ Recognized with the 3 Nference’s Excellence Award for Outstanding Perfor- mance for exceptional contributions and leadership.
⋆ Developed and maintained critical features for real-time backend applications - patient signals and cohort visualizer, processing Electronic Health Record (EHR) data for 6+ million patients and serving major healthcare providers such as Mayo Clinic, Duke Health, and Mercy Health.
⋆ Enhanced text search app infrastructure by redesigning indexes, resulting in a over 40% improvement in real-time search performance despite increasing data vol- ume.
⋆ Scaled applications to support three additional pharmaceutical clients while cutting infrastructure costs by over 15% through Kubernetes-managed Docker de- ployments.

I work at Advanced Numerical Simulations Laboratory (ANSLab), University of British Columbia to generate Unstructured Quad-Dominant Anisotropic Meshes.

I wrote an automatic and fast method to discretize the surface of geometric domains (bodies) in a way that they can be used to generate volume meshes (volumetric discretization).

To know about my work and find out why it is so important, browse the project description.

I worked at Propulsion and Energy Conversion Laboratories, IISc, Bangalore (now at University of Toronto) for one and a half years.

I pursued a highly interdisciplinary project wherein I found correlations between flame propagation and complex network theory. Here, I developed an algorithm to convert spatial flame fronts to complex networks.

Know more about my work in the project descrition here.

Advancing Layer Surface Mesh Development

Discretizing geometry using unstructured meshes is a vital part of any simulation. In my Master's thesis project, I wrote a mesh generator which produces anisotropic surface meshes ideal for boundary layer flow simulations. These meshes act as the foundation for generating hex-dominant volume meshes with complete 3D anisotropy.

I worked at ANSLab, UBC to complete this thesis project. The app was made a part of the in-house mesh generator GRUMMP.

I presented my work at AIAA Scitech Forum 2020. The paper can be found at AIAA Repository - Advancing Layer Surface Mesh Generation. My paper received the Best Student Paper Award - Meshing Visualization and Computational Sciences for the conference.

Basic Viewers for Geometry Data Structures

As a part of Google Summer of Code 2019 project at Computational Geometry Algorithms Library (CGAL), I developed viewers for various geometry data structures.

These viewers are important to quicly visualize the result of an algorithm or to debug code. Viewers were developed for several packages in the CGAL repository including Voronoi Diagram, Periodic Triangulation, Nef Polyhedra and Arrangement 2D.

Find more about my work and the code at GSoC Archive.

Automatic Secondary Motion: Dynamic Kelvinlets

Adding secondary motion to animations is tricky for animators and waiting for another simulation to add such motion is inconvenient. Hence, a solution is to automatically add secondary motion to an object by solving the material elastodynamics equations.

Inspired by the work done at Pixar, I implemented and evaluated the quality of Dynamic Kelvinlets, a analytical technique for real-time physically based animation of virtual elastic materials. Find the implementation on Github.

The implementation details and evaluation of the technique can be found in the report, which was submitted as the part of a course project. Credits to Dave for the ideas and contributions to the project.

Network Structure of Turbulent Premixed Flames

In this interdisciplinary project, I had the opportunity to work on experiments in turbulent combustion as well as algorithm development. The goal of the project was to characterize turbulent premixed flame fronts using complex networks and hence, provide a framework for possible flame interaction and subsequent control.

I developed an algorithm to convert spatial flame front images, obtained via Planar Laser Induced Fluorescence (PLIF) into complex networks. This study produced highly enriched graphical representation of flame fronts and could serve as a framework for characterizing flame-turbulence interactions. The study was published in Chaos: An Interdisciplinary Journal of Non-Linear Science.