I’m a Computer Science student at the University of Toronto working around computer vision, neural graphics, AI systems, and creative developer tools. I like projects where research ideas have to survive contact with real software.
research— neural rendering, 3D avatars, reconstruction, novel-view synthesissystems— local-first software, agent infrastructure, tools that are inspectable and recoverableinterfaces— making technical systems feel playful without making them feel like toys
|
A local-first AI meeting companion that captures conversations, screens and live notes, then turns them into searchable decisions, actions and context. AI SYSTEMS · LOCAL-FIRST · 2026 |
Fast local code radar for coding agents — ranking files, symbols and dependency relationships without putting an LLM in the core search loop. DEVELOPER TOOLS · RUST · AGENTS · 2026 |
|
A privacy-first, self-hosted notification system that calls you back when long-running terminal work needs your attention. SYSTEMS · RUST · SELF-HOSTED · 2026 |
Runtime evidence for agent-authored documentation: inspect the repo, execute declared workflows, capture evidence, track freshness, then safely review and apply docs. AGENTS · DOCUMENTATION · TOOLING · 2026 |
Not a list of every library I have ever installed. This is the stack and set of capabilities that keeps showing up across the things I actually build.
The map is grounded in the repository history — from PyTorch/NeRF experiments and R3F work to Rust code intelligence, Go networking tools, local AI software, browser automation, CI and self-hosted infrastructure.
Repository trail behind the map
| Research + vision | Tiny-NeRF · GaussianAvatar · Music_Avatar · Word-As-Image-HF |
| Systems + agents | agentgrep · shellbell · Re:Meet · quickget · DemoWeave |
| Graphics + interfaces | 3D-Website · NEAT-Unity · Nabokov-s-Cards · ashishajin.com |
| Infra + automation | Compute_Platform · DemoWeave · Re:Meet · shellbell |
A small automatically refreshed window into what is actually moving on my public GitHub. No streak score, no fake proficiency percentages.
Refreshed nightly from the GitHub API by this profile repository.
Vision / graphics. I’m interested in machines that can represent, reconstruct and render the visual world — especially neural rendering and avatars.
Useful AI systems. I care less about bolting a chat box onto things and more about building the execution, retrieval, evaluation and interface layers that make AI tools dependable.
Small tools with sharp edges. CLIs, self-hosted services, weird browser experiments, infrastructure, and things I build because I wanted them to exist.
// still making a mess — just with better abstractions.