xing gao.
Xing Gao on a sunny hillside surrounded by trees

Hello, I’m

Xing Gao.

Backend & Applied AI Engineer

I build AI applications end to end—from React interfaces to reliable Python services.

View Resume View Projects
A little more about me
52model / endpoint configurations
across 8 providers at Vicino
5+PayPal components instrumented
with traces, metrics, and logs
800 → 200msrecommendation API P95
at NaviData AI

A little about me.

I’m a software engineer with 3+ years of experience building backend services and AI applications. I enjoy taking a product from an early idea to something people can use—and making the workflows behind it reliable.

Master of Computer Science
University of Illinois Urbana-Champaign · 2025

Experience

2026 - Present

Software Engineer, Applied AI

Vicino AI

Built multimodal creation workflows from React interfaces to Python services, integrating 52 model and endpoint configurations across 8 providers. Made long-running jobs recover safely through durable state and idempotent retries.

  • Image, video, 3D
  • Durable async jobs
  • Evaluation loops
2021 - 2023

Software Engineer

PayPal

Developed Java backend components for transaction screening and account review. Added tracing, metrics, and logs across 5+ components to diagnose failures between services.

  • 5+ instrumented components
  • Cross-service diagnostics
  • SignalFx and Splunk
2025

Research Assistant, Backend Engineering

UIUC AI Education Platform

Built a knowledge-graph backend for an AI education platform, connecting 125 course artifacts across 28 units. Added LLM-assisted concept extraction and human-reviewed publishing.

  • Neo4j knowledge graph
  • FastAPI and React
  • 125 course artifacts
Earlier

Backend Engineering Internships

Bili and NaviData AI

At Bili, built agent workflows for editing Google Docs and Sheets. At NaviData AI, developed recommendation APIs and reduced P95 latency from 800ms to 200ms.

  • LangGraph
  • FastAPI
  • Azure SQL and Redis

Selected projects

Knowledge graph linking student submissions, assessments, course materials, and concepts

UIUC Research Assistant

AI Research Platform

An education platform that connects course materials, student work, and concepts in a knowledge graph.

My contribution
Built backend workflows for material ingestion, LLM-assisted extraction, human review, and instructor analytics.
Result
Organized 125 course artifacts across 28 units; parallelized extraction from 6 sequential calls into 2 concurrent rounds.

Built with

  • FastAPI
  • React
  • Neo4j
  • Knowledge Graph
  • Concept Extraction
TokenCause diagnostic report preview for local AI coding sessions

Developer tooling

TokenCause

Code

A local tool for understanding where AI coding sessions spend time and tokens.

My contribution
Built a Python CLI that parses session traces and detects retry loops, repeated work, context drift, and expensive tool output.
Result
Produces readable diagnostic reports and dashboards, plus JSON output for automation.

Built with

  • Python
  • Local CLI
  • Trace Parsing
  • HTML Reports
  • JSON Output

Cozad startup project

Moxie

Turns course notes, slides, and PDFs into short playable games.

My contribution
Built the MVP flow for AI authoring, creator approval, publishing, play links, and QR sharing.
Result
A working content-to-game flow with analytics for player choices, endings, and drop-off points. Watch the demo alongside this description.

Built with

  • Next.js
  • TypeScript
  • AI Authoring
  • Vercel KV
  • Analytics
Fluxa procurement war room showing BOM risk scoring and build-risk controls

CacheHacks software track

Fluxa

Demo

A hackathon prototype that helps hardware teams explore procurement and supply risks.

My contribution
Built workflows for editable bills of materials, supply signals, deterministic risk scoring, and AI-generated analyst briefs.
Result
Compares what-if actions and produces an operator-approved execution pack. Explore the linked demo.

Built with

  • React
  • Vite
  • Analyst Briefs
  • GDELT
  • Risk Engine
Leader Follower Follower Log Heartbeat Recovery

Distributed systems coursework

Distributed Systems Suite

A coursework project exploring how a distributed file system behaves when nodes or networks fail.

My contribution
Implemented Raft consensus, leader election, heartbeats, log replication, and recovery behavior.
Result
Tested a 10-node cluster with node failures and network partitions, including two-fault tolerance.

Built with

  • Go
  • Java
  • gRPC
  • Raft
  • Fault Tolerance

Other things I’ve built

Research agent

DeepSeeker

Self-healing deep research agent with planner, web workers, critic-triggered retries, cited reporting, persisted state, and trace-based debugging.

Repo
Agent workflow

Google Docs Integration

LangGraph document generator with parallel section generation, Google Docs creation, OAuth fallback, image/table insertion, and section-by-section editing.

LLM simulation

Simulation Agent

Mesa-based simulation runtime where LLM agents make natural-language decisions, collect behavior data, and generate analysis reports.

Technical toolkit

Languages

  • Java
  • Python
  • Go
  • C++
  • TypeScript
  • JavaScript

Backend Platforms

  • Spring Boot
  • FastAPI
  • Node.js
  • gRPC
  • Hibernate
  • Express.js

Data Systems

  • PostgreSQL
  • MySQL
  • MongoDB
  • Azure SQL
  • Neo4j
  • Redis

AI Workflows

  • LLM Apps
  • Agentic Workflows
  • Multimodal AI
  • Evaluation
  • Prompt Optimization
  • Async Orchestration

Cloud & DevOps

  • AWS
  • GCP
  • Azure
  • Docker
  • CloudWatch
  • Git

Reliability

  • OpenTelemetry
  • Splunk
  • SignalFx
  • Monitoring
  • Microservices
  • System Design

Resume

The one-page version.

A compact version of my experience, projects, and technical stack for recruiter review.

Open Resume Download PDF
@

Email Me

happyxgao@gmail.com

↗

Open For

Backend, AI infrastructure, and applied AI roles.

Send a Message

0 / 5000