I'm a Computer Science Honours student at Monash. I build tokenizers in Rust, distributed backends in Go, transformers in PyTorch. I'm particularly interested in agentic systems and distributed systems engineering.
01 · Grounded
A local-first evidence workspace for students and researchers.
I keep losing things in my own files. A paper I annotated weeks ago, a note I wrote for a past assignment, a half-finished paragraph of my thesis. I can feel the idea is in there somewhere, but finding it takes long enough that I usually give up and rewrite it.
The core idea is not to generate text. We lack an efficient, local retrieval application that can find the passage that already exists, and show us exactly where it came from.
Grounded is my attempt at fixing that. It is a local-first desktop app that indexes everything I study, read, and write (PDFs, lecture slides, markdown notes, thesis drafts) and returns the exact passage for a query, with a citation pointing back to the file and page.
I am building it because I need it. My honours year is a pile of audio deepfake papers, scattered notes, and drafts I keep forgetting I wrote. Nothing I have tried retrieves at the level I want, so I am starting from the bottom: plain lexical search in Rust with Tantivy, SQLite for metadata, a Tauri shell on top. I plan to add semantic retrieval after the core is solid.
I am not building another AI wrapper that can chat with PDF, act as a Zotero replacement or a note app with AI panel stuck on the side. Grounded is only meant to be a local index and evidence surface, so the tools I already use can stay in place.
02 · Projects
I list six projects by how deep I went technically: a tokenizer in Rust, a banking core in Go, a transformer in PyTorch, and more.
03 · Stack
I group these by where they sit in my stack, and I actually use every item in something I've shipped or in my current research.
04 · Education
I finished my undergraduate work in algorithms and data structures, and I'm on honours research in audio deepfake detection.
2026
Monash University · Melbourne, Australia
In progressHonours in partnership with the Australian Federal Police.
Synthetic audio shows up in fraud and impersonation, and a quick listen is not a reliable test. Current deepfake detectors cannot generalize whether a clip is fake or authentic when the clip is short, compressed, or from a system they were never trained on. My thesis aims to extend upon recent breakthroughs in this space by analyzing whether discretized acoustic features help more than continuous ones for detection and for localizing where a clip was faked. I evaluate all major benchmarks, experiment with self-supervised encoders and tokenized speech, and ultimately build a model pipeline to compare how different audio representations perform.
2022 to 2025
Monash University · Melbourne, Australia
Graduated December 2025
