Pinhao Song
中文
Pinhao Song

AI Product Builder · Beijing

AI Product Builder · BIT

I turn AI systems into products people can actually use.

I'm Pinhao Song, an Electronic Information Engineering undergraduate at Beijing Institute of Technology. I build agent workflows, generative media tools, and interactive products—from product definition to deployed demo.

View selected workEmail me
Based in
Beijing, China
Status
Open to product collaborations
Direct
Email · GitHub

Selected work · 01–03

Selected systems, shipped end to end.

Each project begins with a real need, becomes working software, and is documented for human review.

01

Human-reviewed generative media

FashionAI Studio

FashionAI Studio product workflow interface

A six-stage editorial workspace from garment facts and source media to reviewed outputs and a manifest-backed ZIP.

  • Six explicit workflow stages
  • Quality evidence + human selection
  • Original demo pack + manifest ZIP
Case studyLive Mock v7Source
02

Behavior design

TT16 · TradeType 16

TT16 trading-style assessment interface

Twenty realistic decision scenarios become sixteen trading styles and a complete browser-local report.

  • 20 decision scenarios
  • 16 trading styles
  • Local report + export
Case studyTry itSource
03

Agent workflow

StoryWeaver

StoryWeaver long-form fiction workflow

A Claude Code plugin for Chinese fiction projects beyond 300,000 characters, built around repeatable planning and verification.

  • v1.2.3
  • 5 agents · 27 commands
  • 13-dimensional consistency checks
Case studyInstallSource

Archive

More shipped work

04

CatOps

Product simulation

A playable cat ad-operations simulator balancing auction revenue, user experience, retention, and LTV.

LiveSource
05

AutoInsight AI

Evidence workspace

An evidence-first autonomous-driving case review and task-handoff workspace.

LiveSource
06

Personality Escape Station

Interactive world

A personality test, shareable identity card, and explorable vertical pixel room.

LiveSource
07

Narrative Trace

Interactive fiction

A vertical interactive-fiction experience with an AI-assisted creation pipeline.

LiveSource
08

BadGuard

Data product

Four explainable daily notes for observing common A-share technical states without issuing trading instructions.

LiveSource
09

Hip 24-Point Annotation Tool

Research tooling

A local DICOM annotation and review workspace with YOLO-assisted initialization.

Source
10

AI Life Hotpot

Generative experience

A team-built multimodal experience that turns life choices into a personalized hotpot report.

LiveSource

Process

How I build

  1. 01

    Make it playable.

    Start with the smallest complete experience a user can finish by hand.

  2. 02

    Make AI controllable.

    Separate generation, validation, review, and publishing into explicit steps.

  3. 03

    Ship the boundary.

    Treat deployment, data definitions, disclaimers, and failure modes as part of the product.

Selected recognition

Recognition

  • Youth Resonance AwardDouyin AI Creator Program Hackathon
  • Excellence AwardMiraclePlus Vibeathon

About

Building useful, playful, and controllable AI products.

I study Electronic Information Engineering at Beijing Institute of Technology. I begin with a small but complete experience, then connect the data, models, product boundaries, and deployment path needed to make it real.

Now

Agent systems, generative media, and interactive products.

Currently exploring clear human checkpoints, playful interfaces, and reliable paths from prototype to deployment.

Toolbox

Claude Code · Codex · Python · TypeScript · React · FastAPI · Node.js · Cloudflare · GitHub Actions

Build together

Have a product worth shipping?

I'm open to AI product collaborations, open-source work, and conversations about AI-native experiences.

Email meGitHub