Skip to main content
T
Tarun 「道」systems & ai
「制作と検証」 • STUDENT PROJECTS & BUILDS

Projects & Engineering Builds

Hands-on tools built to explore applied machine learning, Japanese & Mandarin language acquisition, and systems engineering.

Complete • OCR + Vision
「視覚識別」
CAM-INPUT: GLYPH_FRAME Example Visual
道
1. 辶 (road/walk)
2. 首 (head)
Meaning: way, path

PolyglotLens AI

Lightweight OCR and dictionary lookup tool tailored for Japanese kanji acquisition. Extracts characters from raw camera clips and matches stroke order animations.

Python OpenCV OCR PyTorch
Complete • SRS Engine
「漢字分解」
RADICAL PARSER Example Visual
休 xiū (rest)
→
亻 person
+
木 tree

HanziFlow

Interactive mnemonic flashcard engine decomposing characters into semantic radicals with automated spaced repetition.

TypeScript React SQLite Tailwind
In Progress • Local SLM
「局所推論」
local-copilot-daemon Example Visual
> load_context("arxiv:2305.18290.pdf")
Context window: 4096 tokens [Offline]
✓ Extracted 3 synthesis flashcards

NeuralNotes

Local-first study copilot leveraging open-weights models running directly on device. Synthesizes technical papers and constructs test checks fully offline.

PyTorch HuggingFace FastAPI
Complete • Web Audio Synth
「音響即読」
AUDIO PHONETIC DRILL Example Visual
MORA: コンピューター Flat Pitch

KatakanaSpeed

Interactive audio-phonetic drill trainer with real-time mora pitch detection. Accelerates rapid foreign-loanword decoding through auditory feedback.

Web Audio API JavaScript CSS Audio UI
In Progress • Knowledge Graph
「漢字連関」
RADICAL GRAPH MAP Example Visual
海 ← 水 → 河

KanjiGraph

Semantic graph visualizer mapping shared radicals, phonological shifts, and semantic divergence between Hanzi and Kanji character trees.

D3.js Python NetworkX
💡
Collaboration Slot

Suggest a Project Idea

Have an idea combining lightweight ML models, speech recognition, or vocabulary graph theory? I'd love to explore building it together.