Experience

The University of Osaka — Yamaguchi Lab

Undergraduate / Graduate Researcher

Research on fast, city-scale prediction of sunlight reflected by Urban Air Mobility (UAM) aircraft, part of the lab's work on urban digital twins.

• Led the method design, simulation (Unreal Engine 5 / C++), model implementation (PyTorch), evaluation, and paper writing.
• Reduced a 10,000-point, five-second evaluation from over 3 hours of ray tracing to 1.62 seconds after one-time city initialization.
• First author at IEEE SmartComp 2026; Encouragement Presentation Award at the IPSJ ITS Workshop.

Outcome: Published at IEEE SmartComp 2026. A journal extension is under review at the Journal of Information Processing, and a study on flight planning for agricultural drones has been accepted at the IPSJ ITS workshop.

The University of Osaka — School of Engineering Science

Teaching Assistant

Teaching assistant for the undergraduate course 基礎工学のための情報学1 (an informatics course in the School of Engineering Science).

Timee, Inc.

Intern (upcoming)

An upcoming one-month internship. Details will be added after the program.

How Television (ハウテレビジョン)

Developer Intern

An individual challenge that combined development with operational decisions, built with Claude Code and a GitHub pull request and CI workflow.

• Placed in the top three of all participants on the challenge KPI.
• Learned to make "verify before acting" a routine: check irreversible costs with data first, and confirm key metrics after each change.

M3, Inc.

Data Analysis Intern

Acting as the data scientist on a fictional case, I developed a data-based promotion strategy and presented it to the team.

• Narrowed several directions down to one proposal and backed it with multiple analyses in Python (pandas, statsmodels, matplotlib), using Claude Code to speed up analysis and slide-making.
• Turned the analysis into Japanese slides for non-specialists, revised them after mid-point mentor feedback, and gave a 7-minute presentation with Q&A.

Outcome: Mentors praised my speed, the multiple analyses behind the proposal, and how clearly the opening slide stated the key message.

Stanby, Inc.

Engineering Intern

The team set out to give job seekers a clearer reason to register. I proposed a feature that automatically applies to popular, fast-closing jobs as soon as they match a member's saved preferences.

• Designed the system architecture and implemented the Python backend and SQLite database for preferences and application workflows.
• Agreed on API interfaces early so the frontend and backend could be built in parallel, and helped integrate the Nuxt frontend.
• Worked out the role and inputs of an AI matching model with the ML member (design stage only).

Outcome: Delivered an operable end-to-end demo within four days (a prototype, not released to production).

PKSHA Technology

Software Engineering Intern

Valuable Slack posts get buried in the message stream. We built a tool that ranks each week's posts by value-aligned reactions and shares them back with LLM-written summaries.

• Proposed the architecture and technology choices — EventBridge schedule → AWS Lambda (Python) → Slack API → Azure AI API → Slack.
• Implemented the backend and API integrations for weekly aggregation, ranking, and summary generation.
• Shared design rationale, reviewed code, and supported teammates with less backend experience.

Outcome: Completed a working tool and presented it at the end of the program.

H&K Inc.

Software Engineering Intern

Built an integration that lets businesses send fincode payment links from HubSpot and syncs payment status back, reducing manual hand-offs and tracking.

• Starting from my supervisor's outline, handled technical research, the backend, authentication, error handling, and API integration with Python, Flask, and AWS Lambda.
• Found a nine-hour timestamp mismatch during integration testing by tracing the different time references used by each service.

Outcome: The project was discontinued at roughly 70–80% completion when the company changed direction. It gave me end-to-end ownership, from research to integration testing.