Xiangyu Shi

01 ~/about

About

How I work, and what I'm like when things get hard.

Different stacks, different problems: industrial telemetry, live captions for deaf students, hotel websites, a behavioral health platform. What I'm good at, in all of them, is landing in a codebase I didn't write, finding the part that's actually broken, and getting the fix shipped.

I stay on it past the merge

Owning a service means the deploy, the incidents, and the ticket that lands three weeks later, not just the pull request.
I go to the data before I form a theory

Root cause usually shows up in telemetry before it shows up in the code. Data is the evidence, so I would rather read logs for ten minutes than guess for an hour.
I write in their style and bring the complexity down

I follow the conventions already in the codebase, then push the parts that are O(n) toward O(log n) or O(1), keeping behavior identical.

02 ~/roles

Roles I'm seeking

Six directions I'd take, with the work behind each. Full tech list under Skills.

Software Engineer

Owning a service end to end — schema, API, deploy — on small teams where that is the job.
Full-Stack Engineer

Database schema through to the interface, shipped as one piece of work.
Back-End Engineer

APIs and services that stay fast, observable, and simple to operate.
AI/ML Engineer

Applied ML and LLM work that ships: computer vision, model pipelines, and agent workflows.
Database Engineer

Schema design, query tuning, and indexes that keep endpoints fast under real traffic.
Front-End Engineer

Interfaces that load fast, hold up at any screen size, and work with a screen reader.

03 ~/education

Education

Rochester Institute of Technology, M.S. Computer Science

M.S. Computer Science · Dec 2025

AI / Machine Learning

Machine Learning & AI Principles · Computer Vision · Python for AI/ML

Data Science & Data Engineering

Big Data Frameworks & Analytics · Database Internals & Storage Engines · Statistical Modeling & Data Mining

Software Engineering & Systems

Object-Oriented Design & Architecture · Core Languages & Ecosystems · Algorithms & Data Structures · Networking & Distributed Systems · Building a Project End to End

Rochester Institute of Technology, B.S. Web & Mobile Computing

B.S. Web & Mobile Computing · May 2023

Front-End & UI/UX Development

Client-Side Web Engineering · UI/UX Design & Prototyping · Digital Visual Assets

Back-End, Databases & Systems

Server-Side Development · Relational & NoSQL Databases · Network & Systems Fundamentals

Mobile Application Development

Native & Multi-Platform Mobile Apps · Mobile Data & Cloud Integration · Shipping a Mobile App

Core Software Development

Software Design & Best Practices · Core Programming Foundations · Applied Mathematics & Statistics · Team Project Management

04 ~/skills

Skills

Everything I've used enough to be useful with.

Programming Languages

~/lang.ts

Python
JavaScript
TypeScript
Java
C#
C++
SQL

AI / ML

~/ai-ml.py

PyTorch
Computer Vision
Deep Learning
Graph Neural Networks
Model Quantization
LLM Integration
Generative AI
LLM fine-tuning
Prompt Engineering
Model Evaluation
Data Preprocessing
TensorFlow
scikit-learn

AI Tools

~/tools.md

Cursor
Claude Code
ChatGPT
GitHub Copilot
Gemini
Agent Workflows

Back End

~/backend.ts

Node.js
Express.js
NestJS
FastAPI
ASP.NET / .NET Core
Spring Boot
REST APIs
GraphQL
Microservices
Distributed Systems
RabbitMQ
Service Bus
WebSockets
JWT

Front End & Mobile

~/frontend.tsx

React.js
Next.js
Angular
Vue.js
HTML5
CSS3
Tailwind CSS
Responsive UI
Accessibility
iOS / Android

Databases & Data

~/db.sql

PostgreSQL
MySQL
Microsoft SQL Server
MongoDB
Cosmos DB
Redis
SQLite
NoSQL
KQL
Schema Design
Indexing & Query Tuning

Cloud & DevOps

~/cloud-devops.yml

AWS
Azure
Azure Functions
Azure Container Apps
Azure Service Bus
Azure Event Hubs
Docker
Linux
Git
GitHub Actions
CI/CD

Testing

~/tests.test.ts

Jest
pytest
xUnit
Unit Testing
E2E Testing
Test Case Design

05 ~/experience

Experience

Where I've shipped production software.

Ortusify LLC

Software Engineer

Jan 2026 – Present

Built new services and APIs from 0 to 1 for SaaS products, internal systems, and hotel websites, full-stack on Next.js, TypeScript, and PostgreSQL.
Cut a data-heavy admin listing from 173ms to 37ms with keyset pagination and covering indexes.
Designed and shipped dozens of high-concurrency REST and GraphQL endpoints with connection pooling and caching, holding p95 latency in the tens of milliseconds in production.
Built the CI/CD on GitHub Actions with Docker deployments, cutting manual release time by over 60%.
Shipped production services in C# / .NET, Java / Spring Boot, Python, and Node, deployed on both AWS (EC2, S3) and Azure.
Cut the PR cycle from two weeks to under three days with Claude Code and GitHub Copilot writing summaries and flagging what needed discussion.
Designed front ends with AI tools, writing prompts tuned to each product's intended look and feel.
Used AI code review to catch risky code and raise quality, and leaned on it to generate test cases and learn unfamiliar patterns.

Next.js · TypeScript · PostgreSQL · C# · .NET · Java · Spring Boot · Python · AWS · Azure

Center on Access Technology, RIT/NTID

Software Engineer (Part-Time)

Aug 2022 – Dec 2024

Built a real-time speech-to-text captioning app in C# and .NET Core, used by 200+ deaf and hard-of-hearing students across 50+ meetings and events.
Led the design of the async message pipeline: RabbitMQ with persistent WebSockets, plus acknowledgment and reconnection strategies that eliminated packet loss in live captioning.
Profiled the message-processing chain and removed latency bottlenecks, cutting end-to-end caption latency by nearly 40%.
Ran accessibility test cycles for PopSign, an ASL learning app from the Georgia Tech and Google teams.
Used GitHub Copilot for code generation, review, pull requests, and test cases.

C# · .NET Core · RabbitMQ · WebSockets · Accessibility

TigerChat · PopSign · Center on Access Technology

Ecolab

Software Engineer Intern

Jan 2024 – Aug 2024

Built serverless telemetry pipelines in C# / .NET Core on Azure Functions, Event Hubs, and Service Bus, handling 5M+ messages a day from a 100,000-device fleet worldwide.
Optimized Cosmos DB queries over 1M+ production records with key indexes, query rewrites, and snapshot tables.
Turned hardware contracts into stable, versioned API schemas exposed over REST.
Deployed the services as containers on Azure Container Apps, holding xUnit coverage above 90% as the schemas evolved.
Investigated field device failures in Azure Data Explorer (KQL) to find root causes and resolve production incidents.
Used GitHub Copilot for code review, PR summaries, test cases, and getting up to speed on unfamiliar code.

C# · .NET Core · xUnit · Azure Functions · Event Hubs · Service Bus · Cosmos DB · Container Apps · KQL

CHESS Health

Software Developer Intern

Jan 2022 – Aug 2022

Built backend services in Node.js with Express.js, and in ASP.NET Web API, for a behavioral health platform.
Delivered dozens of REST endpoints across MySQL, MS SQL Server, and MongoDB, with Jest covering the suite.
Shipped front-end features in Angular, React, and Vue.js.

Node.js · Express.js · Jest · ASP.NET · MySQL · MongoDB

06 ~/projects

Selected projects

Two I'd show you first, and three more underneath.

~/papertab.ts

Papertab

Full-stack · live app

Splitting the bill "evenly" is lazy and usually wrong. Add a receipt, check off exactly what each person ordered, and it works out individual tax and tip, tracks who's paid, and shows who still owes what, down to the cent. An AI ingestion path reads the receipt photo and extracts the merchant, line items, tax, and tip into structured JSON.

Next.js · TypeScript · Tailwind CSS · RESTful APIs · PostgreSQL · Redis · AI Integration

Visit Papertab

~/handwriting-gcn.py

AI-Powered Chinese Character Recognition & Learning System

M.S. capstone · Jan 2025 – Dec 2025

For my M.S. capstone I built a system that reads Chinese handwriting. Each character is modeled as a stroke graph, and a Graph Convolutional Network trained in PyTorch Geometric hit 99.0% top-1 accuracy across 2,000 character classes. I quantized the model and tuned inference to run locally on an M1 Mac, reaching 11 characters a second at 108ms p95 from raw Apple Pencil input.

Python · PyTorch Geometric · Graph Neural Networks · Computer Vision

99.0% top-1 · 2,000 classes

More projects

~/newworld.ts

New World — Board Game Competition Platform

B.S. capstone · Aug 2022 – May 2023

A full-stack platform for board-game competitions that our capstone team built on Next.js and NestJS. I designed dozens of PostgreSQL-backed REST endpoints with JWT auth and role-based permissions, and it carried a hundred concurrent users through a live tournament.

Next.js · NestJS · PostgreSQL

~/http-server.cpp

multiple-thread-http-server

C++17 · Linux

An HTTP/1.1 server written from scratch in C++17 with a single-Reactor thread pool, edge-triggered epoll dispatch, hand-rolled HTTP parsing, and async logging. Mostly my way of understanding what actually sits underneath frameworks.

epoll · Reactor pattern

GitHub

~/research-agent.py

agentic-research-assistant

Python · AI agents

A CLI agent that keeps going until it actually answers the question. It searches the web, does exact math, and keeps notes in a persistent notebook across steps.

LLM integration · CLI