About Me#

Tokyo, Japan

Email: ykwang224[AT]gmail.com


Education#

The University of Tokyo · 東京大学
M.S. Candidate, Information & Communication Engineering — Oct. 2025 – Sep. 2027 (Expected)

  • Suzumura Laboratory. Researching Vision-Language-Action (VLA) models and human-robot collaboration benchmarks for scientific laboratory environments.

Shanghai University · 上海大学
B.E., Electronic Information Engineering — Sep. 2021 – Jul. 2025


Projects#

KotobaLab#

Personal · SwiftUI · GRDB 7 · SQLite · SwiftData · Swift Testing · iOS 18+
Apr. 2026 – Present · Tokyo

Local-first Japanese dictionary app built with SwiftUI, MVVM, and a Clean Architecture–inspired layered design. Currently in local development.

  • Built a local-first dictionary app with a 293K-entry, 52 MB SQLite database generated by a Python pipeline, bundled as an app resource and queried through GRDB DatabaseQueue.
  • Optimized prefix search with PRAGMA case_sensitive_like=ON, improving benchmarked queries from ~16 ms table scans to ~0.03 ms multi-index lookups.
  • Designed a four-layer architecture (AppDependencies → Scene → Store → View) following MVVM and Clean Architecture principles; applied the Repository pattern with protocol-based abstractions and dependency injection to decouple Domain logic from UI and storage frameworks.
  • Combined GRDB-managed SQLite for read-heavy reference data with SwiftData for user state; wrote 13 unit tests with Swift Testing covering domain use cases and repository behavior.
  • Roadmap: SQLite FTS5 full-text search, actor-isolated repositories under Swift Concurrency, TestFlight beta.
shiinayane
/
KotobaLab
Waiting for api.github.com...
00K
0K
0K
Waiting...

LabMate#

Suzumura Lab · Python · Isaac Sim · VLA Models
Dec. 2025 – Present · Tokyo

Simulation-first benchmark for human-robot collaboration in scientific laboratory environments.

  • Extending the LabUtopia simulator into a benchmark for natural-language human-robot collaboration in scientific laboratories, focusing on clarification, safety-aware refusal, grounding, and structured action logging.
  • Designing an evaluation framework to compare planner architectures (LLM, VLA, and hybrid systems) on multi-step laboratory protocols; sim-first phase planned for 6+ months before hardware integration.
  • Coordinating with 3 procured robotic platforms for future hardware integration; prioritizing rigorous evaluation methodology over demonstration-driven research.

Experience#

GlucoPI#

B.E. Graduation Project · Advisor: Qi Zhang, PhD, Professor
Feb. 2025 – Jun. 2025 · Shanghai

WeChat Mini Program for diabetes management, doctor-patient communication, and glucose prediction.

  • Built an end-to-end diabetes self-tracking app with WeChat Mini Program frontend, FastAPI backend, MySQL / MongoDB storage, and WebSocket doctor-patient chat.
  • Implemented login and permissions, glucose / diet / insulin records, doctor-patient pairing, real-time messaging, and trend charts.
  • Added short-horizon glucose prediction in PyTorch by reproducing the GluPred pipeline and running experiments on OhioT1DM.
  • Integrated a lightweight LLM Q&A feature for non-diagnostic health guidance.
shiinayane
/
glucopi
Waiting for api.github.com...
00K
0K
0K
Waiting...

Arrived or Not#

Co-Leader · Advisor: Qi Zhang, PhD, Professor
Dec. 2023 – Dec. 2024 · Shanghai

Machine-vision teaching platform for classroom attendance and course management.

  • Implemented real-time face-recognition attendance with RetinaFace detection and ArcFace embedding-based identity matching.
  • Built teacher / student mobile workflows for class setup, attendance records, and basic engagement statistics.
  • Developed a Flutter client and FastAPI service with a clear frontend-backend split.
shiinayane
/
Arrived-or-Not-Frontend
Waiting for api.github.com...
00K
0K
0K
Waiting...
shiinayane
/
Arrived-or-Not-Backend
Waiting for api.github.com...
00K
0K
0K
Waiting...

Interests#

  • Product-First Engineering — Building products that solve real problems; treating iOS, backend, and full-stack as means rather than ends.
  • iOS Native Architecture — Local-first design with clean Domain / UI separation; currently focused on shipping native iOS apps as the most direct path to product impact.
  • Research-to-Product — Bridging VLA / LLM research from graduate work to consumer mobile experiences.

Skills#

  • iOS / Apple Platforms — Swift, SwiftUI, GRDB, SQLite, SwiftData, Observation, Swift Concurrency, Swift Testing
  • Architecture & Design — MVVM, Clean Architecture, Repository Pattern, Dependency Injection, Protocol-Oriented Programming
  • Backend / Web — Python (FastAPI), JavaScript
  • ML / Research — PyTorch, VLA Models, Isaac Sim
  • Tools / Platforms — Git, Docker, Xcode, VS Code, macOS, Linux

Languages#

  • Japanese — Professional working (JLPT N1 121/180)
  • English — Professional (TOEFL 97/120)
  • Chinese — Native