Ashtan Mistal
I'd also recommend visiting my LinkedIn as it is updated more frequently than this page. Feel free to reach out there if you want to chat!
Work Experience
Software Engineer (C/C++)
Polyga, Inc.
Jun 2024 - Present
- Point set registration: Overhauled the point-to-plane ICP alignment algorithms, making major improvements in robustness, ease of use, and accuracy across a wide range of data for both automatic and real time alignment.
- 3D color processing tools: Designed and built a layer-based, non-destructive color processing suite with support for filters, color enhancement, blend modes, and and vector field processing, bringing Photoshop-like capabilities to 3D objects.
- Expanded geometry tools: Precise 3D measurement capabilities, best-fit shape primitive calculation, and support for cylinders, circles, and precise shape editing and creation tools.
Applied Physics Software Developer
Mitra Biotechnologies
Dec 2023 - May 2024
- Rapid prototyping and computational modelling of lateral flow assays through a full-stack web application, using TypeScript and Python with React, Django, PostgreSQL, and Docker.
- Utilized OpenFOAM in C++ for fluid dynamics simulation in paper-based porous media, using GMSH for quad mesh creation and optimization from user-provided GeoJSON polygon data. Designed a project management backend, database, and RESTful API for the platform in Django.
Undergraduate Teaching Assistant
UBC Department of Computer Science
Sep 2021 - Dec 2023
- Facilitated student comprehension of computer science concepts and programming techniques through proactive engagement during weekly office hours and lab sessions.
- Engaged closely with fellow teaching assistants and professors to create a dynamic and supportive learning environment, ensuring students received comprehensive guidance in a fast-paced educational setting.
- Received consistently positive performance evaluations from students, highlighting the impact of my teaching methods in driving their academic success and fostering their enthusiasm for computer science.
- Employed effective problem-solving and debugging skills to assist students in resolving conceptual roadblocks, solidifying learning outcomes and increasing confidence in tackling technical issues.
Residence Advisor
UBC Student Housing and Community Services
Aug 2021 - Dec 2023
- Skillfully managed conflicts by providing a safe and supportive space for residents to address issues, employing active listening and mediation techniques to facilitate constructive resolutions.
- Helped strengthen the residence experience through supporting safety and security, personal development, and wellness through a peer-to-peer approach.
- Coordinated with a team to develop and implement a variety of programs and events to foster a sense of community and belonging among residents.
Personal Interests
Alongside my academic and professional endeavors, I engage in a variety of extracurricular activities that enrich my life and work. My passion for outdoor activities like hiking allow me to explore the stunning landscapes of British Columbia and beyond. I've also been rock climbing for as long as I can remember.My creative side finds expression in music production, an endeavour I've explored for about 10 years now, and a little bit of digital art. I find both are excellent outlets to express myself while also learning new skills - and often finding inspiration for technical projects along the way.
Building UBC in Minecraft
Crafted a 1:1 digital clone of the UBC campus in Minecraft by designing a multilayered 3D surface reconstruction pipeline.
- Transformed LiDAR and wayfinding data into a 3D model of the campus using geometric modelling and computer vision techniques in Python, generating detailed features such as buildings, trees, roads, and water bodies from the data.
- Presented a technical demo at CVPR 2024, selected as a top performer and receiving an Honorable Mention for Best Demo.
- Engaged with 500+ players and had a social media outreach of 130k+ views.
Forest Friends
- Implemented a research-based machine learning pipeline to cluster trees, using mean shift clustering and vertical strata analysis, adapting the theoretical model to retain voxel accuracy and reducing runtime by over 90%.
- Trained a PointNet++ model in PyTorch to classify individual trees into broad biological families with 89.5% accuracy.
- Automated data labelling using GPT-3 and GBIF APIs to gather and aggregate species information.
LiDAR-DenseSeg
This project aims to improve downstream voxelization and mesh reconstruction tasks through a deep learning pipeline focused on segmenting, densifying, and planar flattening of airborne LiDAR data. A PointNet++ semantic segmentation model was trained to learn building structures in the data, achieving an evaluation accuracy of 94.1% and a mean IoU of 0.89. Following segmentation, the point clouds were densified by developing a recursive median split algorithm that adapts an existing model to work with arbitrary-sized point clouds. Planar flattening was proposed to further refine the process, reducing noise and enhancing voxelization quality by ensuring that neighbouring points are brought to the same voxel.
PET Image Reconstruction: Fourier Filter Analysis
This project involved evaluating the impact of Fourier transform-based filter functions on the quality of PET image reconstruction, beginning by collecting data of a pair of radioactive sources with which to reconstruct. It entailed designing computational experiments to assess filter performance, utilizing image quality metrics as benchmarks, and writing a comprehensive research manuscript detailing the methodology and potential applications within medical imaging.
Jotmark
A web application designed during winter break that allows users to create and edit notes using Markdown syntax. The UI is designed for ease of use and efficiency, inspired by the layout of Google Keep. Enables input of LaTeX math equations, rendered live using katex. Export options for Markdown or LaTeX files. Built using Angular and Material Design components, with a Typescript and Node.js backend. A useful tool for note-taking and document writing in a STEM academic environment.
Teaching Computer Graphics
This website is the culmination of my PHYS 420 project, featuring hands-on and virtual demonstrations to teach secondary school students the basics of shading in computer graphics and the theoretical basis for ray tracing. Includes a presentation on shading basics and ray tracing theory for beginners, with virtual demonstrations coded in WebGL and hands-on demonstrations using light to visualize shading on surfaces.
Numerical Methods in Radiation Room Shielding
Researched Monte Carlo numerical approximation methods for particle interactions to determine the shielding effectiveness of a treatment room in a radiation therapy clinic. Focused on staff safety regarding secondary and tertiary radiation exposure. Discussed practicality and limitations of these methods, including hardware constraints and accuracy of voxel-based geometric modeling of the room. Compiled into a poster for the PHYS 404 class of 2022 at UBC.
Monophonic Synthesizer
Designed and built a monophonic synthesizer using the MSP 430 microprocessor for PHYS 319. This project combined digital electronics, microcontroller knowledge, and audio synthesis techniques. Custom firmware for the MSP 430 was created to generate and manipulate audio signals in real-time. Enhanced understanding of circuitry, microcontroller programming, and audio synthesis, with a button-based user interface on a breadboard for synthesizer control.
Rexfro
Rexfro (Regex Find and Replace Operator) is a Java Swing application that enables users to perform multiple regex find-and-replace operations on text documents simultaneously. This tool streamlines document conversion and translation, providing efficiency in workflows where regular expressions are utilized. With sufficient "dictionaries" of find-and-replace operations, Rexfro can be used to convert documents between different formats, such as Markdown to LaTeX, or to remove unwanted text or encoding errors from documents.