chu1pc
JA

Portfolio

Photo of Tadashi (CHU)

Tadashi (CHU)

Full-stack Engineer

Studying CS at the University of Aizu while learning web app development and ML/DL at 2WINS.

Projects Contact
Scroll

01

  • 2026 Solo project

    Kikitoru

    A web app that turns meeting audio into minutes on its own. AWS Transcribe handles transcription and speaker separation, Gemini condenses the result into structured Markdown, and pgvector makes past minutes searchable in plain language. I built all of it solo, down to Google OAuth, rate limiting, Alembic migrations and CI on GitHub Actions.

    • FastAPI
    • React 19
    • PostgreSQL
    • pgvector
    • AWS Transcribe
    • Gemini
    • Docker
    View on GitHub
  • 2026 Team project (Team1)

    PR TIMES Hackathon 2026

    An AI that invents the PR stories a company has not had yet. For companies coming back from a dormant stretch, it settles on an event worth announcing through a back-and-forth with an LLM and a short interview. We tested the core hypothesis against the dataset we were given, then shipped it to AWS with an OIDC-based CI/CD pipeline.

    • FastAPI
    • React
    • TypeScript
    • AWS (ECS, RDS, WAF)
    • OpenAI API
    • kubb
    View on GitHub
  • 2025 Team project

    DayLog

    An in-house tool for tracking hours and writing the daily report for you. Linear issues sync in over webhooks, and a weekly calendar shows where the time actually went. I owned features such as personal task creation and team switching.

    • Next.js
    • TypeScript
    • PostgreSQL
    • Linear API
    • Google Sheets API
    View on GitHub
  • 2025 Solo project

    "Deep Learning from Scratch" reimplementation

    My own take on the book "Deep Learning from Scratch". Everything from backpropagation to CNNs, built with nothing but NumPy.

    • Python
    • NumPy
    View on GitHub

02

Studying computer science at the University of Aizu, which I entered a year early through the early-entrance program, while interning at 2WINS. There I implement APIs and the data layer, integrate ML and LLMs into real applications, and build the React frontend.

What drives all of it is that I simply like learning. I reimplement papers and books until the ideas are mine rather than quoted. At work I care most about the parts nobody puts in a demo — bugs that only appear under load, caching, anything asynchronous — because those are what decide whether the thing holds up.

Fukushima, Japan ・ University of Aizu

03

  1. 2025.04 — Present

    Backend & ML Engineering Intern

    2WINS, Inc.

    Backend development for an AI product. I integrate LLMs and DL models into the application with strict typing, and run it in production.

    • Backend APIs in Python with FastAPI and Pydantic, with LLMs and DL models integrated under strict typing
    • Database PostgreSQL schema design and migrations with SQLModel and Alembic
    • Infra Deployment and operation on AWS with Docker, ECS, RDS, and S3
    • Frontend Screens in React and Svelte, connected to the backend APIs
    • Quality Faster responses with async processing and caching, tested with pytest and Playwright
  2. 2026.09 Short-term

    Short-term Internship (5 days)

    Progate

  3. 2026.09 Short-term

    Short-term Internship (2 days)

    M3

  4. 2026.08 Short-term

    PR TIMES Hackathon 2026 Summer

    PR TIMES

    Built an LLM-powered AI that helps plan PR campaigns. I worked on it from requirements through to the AWS deployment.

  5. 2025.10 Short-term

    Short-term Internship (5 days)

    AVILEN

  6. 2024.04 — Present

    B.Sc. Computer Science & Engineering

    University of Aizu

    Studying CS with a focus on machine learning, deep learning and data science. Admitted a year early through the early-entrance program (飛び入学).

04

What I can do

  • Full-stack product features.

    I implement features end to end, across infra, frontend, and backend.

  • Backend services & APIs.

    Async APIs on FastAPI + PostgreSQL — SQLModel, Alembic migrations, auth, pytest / Playwright coverage, and Docker / AWS deploys.

  • LLM-powered features.

    Prompt design, structured output, and pipeline work with LangChain and commercial LLM APIs.

  • ML / DL from scratch.

    Transformers, CNNs, and classical baselines in PyTorch and NumPy — built from the paper up, not just wrapped.

Stack I work with

Languages
  • Python Python python.org General-purpose language used for backend and ML.
  • TypeScript TypeScript typescriptlang.org Typed superset of JavaScript used across the frontend.
  • JavaScript JavaScript developer.mozilla.org The standard scripting language of the web.
  • SQL SQL iso.org Query language for relational databases.
Frontend
  • React React react.dev JavaScript library for building UIs.
  • Svelte Svelte svelte.dev Compiler-based UI framework.
  • Astro Astro astro.build Content-focused web framework powering this site.
  • Tailwind CSS Tailwind CSS tailwindcss.com Utility-first CSS framework.
  • Vite Vite vite.dev Fast frontend build tool.
  • Zod Zod zod.dev TypeScript-first schema validation library.
  • kubb kubb kubb.dev Generates Zod schemas and clients from OpenAPI.
Backend
  • FastAPI FastAPI fastapi.tiangolo.com Async-first Python web API framework.
  • SQLModel SQLModel sqlmodel.tiangolo.com ORM combining SQLAlchemy and Pydantic.
  • SQLAlchemy SQLAlchemy sqlalchemy.org Python SQL toolkit and ORM.
  • Alembic Alembic alembic.sqlalchemy.org Database migration tool for SQLAlchemy.
  • PostgreSQL PostgreSQL postgresql.org Open-source relational database.
  • pgvector pgvector github.com Vector similarity extension for PostgreSQL.
  • LangChain LangChain langchain.com Framework for building LLM applications.
  • OpenAI API OpenAI API platform.openai.com API for using OpenAI models.
  • Gemini API Gemini API ai.google.dev API for using Google Gemini models.
  • boto3 boto3 boto3.amazonaws.com Python SDK for AWS.
ML / DL
  • PyTorch PyTorch pytorch.org Deep learning framework.
  • NumPy NumPy numpy.org Numerical computing library for Python.
  • scikit-learn scikit-learn scikit-learn.org Classical machine learning library.
  • OpenCV OpenCV opencv.org Computer vision library.
  • transformers transformers huggingface.co Hugging Face library for pretrained models.
  • unsloth unsloth unsloth.ai Library for fast LLM fine-tuning.
  • Reinforcement Learning Reinforcement Learning spinningup.openai.com OpenAI's educational resource on RL.
Infra & Tools
  • Docker Docker docker.com Containerization platform.
  • AWS AWS aws.amazon.com Amazon's cloud platform.
  • GitHub Actions GitHub Actions github.com CI/CD built into GitHub.
  • Linux Linux kernel.org Open-source OS kernel.
  • uv uv docs.astral.sh Fast Python package manager.
  • Bun Bun bun.sh Fast JavaScript runtime and toolchain.
  • pytest pytest docs.pytest.org Python testing framework.
  • Playwright Playwright playwright.dev Browser automation and testing tool.
  • Git Git git-scm.com Distributed version control system.

Kaggle

Top 2% of 10,148 teams on the Kaggriculture public leaderboard (checked October 19, 2026).

View Kaggle profile View leaderboard

05

Open to internships, research collaboration, and freelance work.