A study of renewable energy production technologies and arctic building techniques to determine the viability of building a power-generating expansion to POLAR Knowledge Canada's High Arctic Research Station in Cambridge Bay, Nunavut. This was in collaboration with, and presented to, representatives from POLAR Knowledge Canada and the University of Ottawa's Civil Engineering Department.
Autumn 2017A shallow raft-foundation design for a proposed eight-story condominium at 231 Cobourg Ave, Ottawa, Ontario. The site at Cobourg Avenue is unique in that beneath a thin layer of sand rests a deep layer of Leda Clay close to the water-table. Leda Clay is known to liquify under pressure when saturated, so increased safety measures were considered in the design.
December 2017A design for a waste-to-energy plant to be built in the High Arctic. This building was designed to reduce Cambridge Bay's ecological footprint while meeting its growing energy needs. With the knowledge gained from the earlier conducted feasibility study, a light-weight building was designed which would sit a few feet above the permafrost on socketed piles. This building also featured state-of-the-art insulation materials. This report was presented to professors in the University of Ottawa's Civil Engineering Department.
Winter 2018A pile foundation design for a proposed development at Tenth Line, Ottawa, Ontario. Iterative design was used to develop a suitable foundation in primarily sand. Site data was determined from borehole logs taken on site.
March 2020A pipeline to create segmented data in several prominent segmented image formats: YOLO and COCO for example. Virtual rooms were procedurally generated in blender and sequences of images, with corresponding segments for objects and walls, were generated. Three-dimensional vertices defining objects in-frame were projected onto the image plane and reduced to only those necessary to express the object's silhouette. Then the segmented data is converted to the desired format. The aim was to determine if a neural network could learn from computer-generated graphics.
Autumn 2020A quick investigation into how space could be parametrized, partly undertaken as a part of Computer Generated Training Data. A design pattern which related pseudorandom elements to collective expression was coded such that random collections of spaces following this pattern could be generated.
Autumn 2020An investigation into the temporal variation of entropic signals, as revealed by time windowing its wavelet transform. A generative adversarial neural network was then developed which sought to learn how the wavelet transform of a signal varied with time and would therefore be able to extrapolate it forward.
Winter 2021An overview of an AI designed to trade cryptocurrency. Modelling financial markets as a stochastic process, the proposed AI implements generative adversarial networks to make reasonable price predictions. Wavelets are also implemented in order to reveal lower frequency trends. Implicit in this design is the cyclical nature of seemingly random events.
Winter 2022An exploration of AI's role in creative work. Primarily consisting of practical analysis, several different AI architectures were studied to obtain a deeper understanding of AI's role in creative space.
Summer 2022A research report on using a meta-learning generative adversarial network to extrapolate nonlinear stochastic systems, with a case study focused on forecasting volatile asset prices for algorithmic trading.
Winter 2023A code-oriented study of the Biondi-Malanga Doppler tomography method, covering literal reproduction, required engineering adaptations, and performance analysis on CPU and CUDA backends.
Spring 2026