Junyu Yao · Rice MCS · UT Austin Physics

AI Product & Platform Engineer

I build production-grade AI workflows, backend systems, and data platform tools across model integration, Firestore-scale migration, microservices, user asset systems, and operational dashboards.

View workResumeGitHub

20+ AI models

Generation workflows across about 12 providers.

922M documents

Checkpointed Firestore migration with verification.

437K calls/day

Modal/FastAPI service workload at production scale.

/ Featured Work

Selected engineering projects

A tighter portfolio focused on production AI platform work, backend/full-stack engineering, and real operational tools. Smaller experiments are kept separate so the main signal stays clear.

OpenArt AI Platform Work screenshot
Production AI PlatformCase Study

OpenArt AI Platform Work

Production AIGC platform work across model integrations, Projects & Folders, Modal/FastAPI services, analytics, and large-scale Firestore migration.

  • Integrated 20+ generative AI models across about 12 vendors and contributed workflows driving about 10.7M generations in 90 days.
  • Built and operated Modal/FastAPI services including a thumbnail service handling about 437K calls/day.
  • Designed and ran a checkpointed Firestore migration across about 922M documents with throttle control, verification, and rollback support.
TypeScriptPythonFastAPIModalFirestoreGCPLaunchDarkly
Read case study
StoryAI screenshot
AI Product PlatformPrivate

StoryAI

AI narrative reflection platform with participant sessions, streaming generation, survey workflows, prompt configuration, and admin monitoring.

  • Session-based AI workflow with persistent memory and variant-specific participant flows.
  • Streaming generation, prompt configuration, admin review, and survey completion tracking.
  • Designed as a research workflow platform rather than a medical or therapy product.
Next.jsFastAPIMongoDBSSEDocker
Private
SnapLink screenshot
Backend / Full-stackIn Rebuild

SnapLink

URL management platform focused on short links, custom aliases, expiration rules, user flows, and click analytics.

  • Full-stack URL shortening workflow with redirect handling and link lifecycle rules.
  • Backend-heavy project for API design, analytics, authentication, and data modeling.
  • Currently being cleaned up and rebranded from its class-project repository name.
Spring BootVueJavaCloud DatastoreREST APIs
Read case studyGitHub
GalaGate screenshot
Event OperationsIn Rebuild

GalaGate

Event check-in and lottery operations platform built around UTCSSA gala workflows serving 1,400+ guests across four large events.

  • Combines attendee check-in, lottery eligibility, and event operations into one system.
  • Evolved from real UTCSSA event tooling rather than a standalone animation demo.
  • Used in event operations connected to 1,400+ guests and $2K+ advertising revenue.
Next.jsNuxtVueFlaskDocker
Read case studyGitHub
UTCSSA Platform screenshot
Community ProductLive

UTCSSA Platform

Community web ecosystem for UT Austin CSSA, including the official website, forum, guide, and event-facing tools used by real student communities.

  • Supported real community traffic across official pages, forum content, newcomer guides, and event tools.
  • Forum grew to 1.7K+ registered users, 2.3K+ topics, and 3.7K+ replies.
  • Demonstrates long-running product ownership beyond one-off prototypes.
ReactNext.jsNode.jsDiscourseVercel
Read case studyWebsiteForum

/ Engineering Focus

The through-line is turning fast-moving product needs into stable, observable systems.

AI product systems

Multi-provider model integrations, generation workflows, asset organization, rollout control, and product analytics.

Backend foundations

FastAPI, Spring Boot, .NET, Firestore, data models, lifecycle rules, analytics, reliability, and deployment hygiene.

Operational tools

Migration tooling, admin workflows, check-in systems, dashboards, and systems used by real operators.

Full-stack delivery

Frontend, backend, data, rollout, observability, and documentation carried as one product surface.

/ About

Production AI experience, backend instincts, full-stack execution.

My strongest proof points sit at the boundary between AI product execution and platform engineering: model-backed workflows, backend services, data migration, analytics, and operator-facing tools.

Junyu Yao

I'm Junyu Yao, a Master of Computer Science graduate from Rice University with a Physics and Business background from UT Austin.

  • OpenArt: integrated 20+ generative AI models, contributed workflows driving about 10.7M generations in 90 days, built Modal/FastAPI services, and helped execute a 922M-document Firestore migration.
  • Dover: built a .NET 8/Microsoft Trill complex event-processing engine integrated with Azure IoT Hub, with cloud cost and subscription revenue impact.
  • Systems toolkit: TypeScript, Python, Java, C#, React, Next.js, FastAPI, Spring Boot, .NET, Firestore/GCP, Modal, Azure, MongoDB, PostgreSQL, Docker, RAG, LangChain, and FAISS.
  • Current portfolio focus: using public-safe case studies for OpenArt, SnapLink, GalaGate, and UTCSSA to show production AI work, backend design, and real community product delivery.

/ Experiments & Archive

Smaller work kept in context

These projects remain useful background, but they do not compete with the main AI/backend/platform story.

Emotional AI

Private

Research-flow prototype for multi-agent group chat, participant surveys, and memory-backed AI interactions.

YTubeDB

Live

YouTube video database and management platform with search, tagging, and content organization workflows.