Machine-readable content: Source catalog · Complete portfolio Markdown · Content index · This page as Markdown.
Aasman Bashyal
Research & AI Architect at Angelswing, Remote.
Angelswing is based in Seoul, South Korea.
I am a Research and AI Architect at Angelswing. I work on problems in computer vision, AI, GIS and deep learning.
Selected work
- Automatic Polygon Drawing
Computer vision
Surveyors can outline stockpiles directly on a drone map. I used weight quantization, a technique from TinyML, to shrink existing segmentation models for local inference with WebAssembly.
- Structure from Motion (SfM)
3D graphics
An SfM pipeline covering feature detection, camera pose estimation, triangulation, and 3D reconstruction.
- 3D Terrain Editing
3D graphics
Editing 3D terrain and flattening selected surfaces into 2D while preserving texture.
- Gaussian Splatting
3D graphics
Rendering novel views with Gaussian Splatting, with the mathematics behind the technique.
Writing
- When weight pruning helps TinyML
· TinyML / model-compression / deep-learning
How weight pruning shapes TinyML models, and where hardware constraints limit the gains.
- When low-pass filtering helps graph neural networks
· GNN / deep-learning
Why low-pass filtering helps graph neural networks, and when removing high-frequency signals can lose useful information.