Software engineering · AI development

Hi, I'm Shenyi Zhang.
I build AI-powered
software.

Software engineer building LLM integrations, agent tools and AI applications, backed by production experience in distributed systems.

Explore my work GitHub
Shenyi Zhang

About me

AI development.
Software engineering.

I'm a software engineer at TikTok USDS JV, working in an SRE role with a focus on hands-on development across AI applications, backend services and infrastructure.

My recent work includes LLM gateway integrations, coding-agent tooling, MCP services and AI-assisted incident workflows.

I also develop backend systems for infrastructure management platforms, including APIs and workflow engines for database provisioning, lifecycle management and capacity orchestration.

Previously at Coles Group, I worked on retail APIs and transaction-processing services. I also have a research background in graph algorithms and fuel-efficient routing.

Connect on LinkedIn

Selected work

Projects & research

AI applications, agent tooling and backend systems — from my current work at TikTok to earlier engineering and research.

TikTok USDS JV

AI development & backend engineering

TikTok USDS JV · LLM integrations

Multi-provider LLM gateway

Contributed regional integrations to an enterprise LLM gateway, helping applications work across model providers through a common interface.

  • Implemented protocol conversion for model requests, responses and streaming output.
  • Contributed asynchronous processing for usage, health and trace data.
  • Adapted the management console for model onboarding, routing configuration and trace visibility.

LLM APIs · Streaming · Asynchronous processing

TikTok USDS JV · Agent development

MCP services for AI agents

Built two MCP services from the ground up, turning manual operational tasks into reusable tools that agents can call.

  • Integrated existing service capabilities through Model Context Protocol.
  • Exposed recurring engineering tasks as agent-callable tools.
  • Made these capabilities reusable across agent-assisted workflows.

MCP · Agent tools · Service integration

TikTok USDS JV · AI applications

LLM-powered incident assistant

Contributed to an internal AI assistant that collects incident context and generates summaries for engineer handoffs.

  • Developed data-collection flows, commands, interactive cards and contextual responses.
  • Contributed incident-context and root-cause summarization for handoff reports.
  • The handoff-report feature was released and used in production.

LLM applications · Data integration · Summarization

TikTok USDS JV · Developer tooling

Coding-agent integrations

Contributed developer tooling around the LLM gateway so engineers could configure coding agents for their environment.

  • Worked on one-click installation and configuration for coding tools.
  • Contributed model discovery and MCP configuration support.
  • Connected coding-agent setup with the gateway integration work.

Coding agents · Developer experience · LLM integration

TikTok USDS JV · Backend development

Database workflow automation

Contributed backend workflows for database provisioning and lifecycle management, replacing repeated manual steps with reusable software.

  • Developed scaling and provisioning capabilities with resource-pool selection.
  • Added approval and execution tracking, plus replica-removal and consistency-remediation workflows.
  • Used reusable templates to separate shared behavior from region-specific execution logic.

Backend services · Workflow design · Database automation

TikTok USDS JV · Distributed systems

Capacity automation engine

Built a state-machine-driven capacity optimization system, translating manual procedures into scheduled, approval-gated execution.

  • Implemented a scheduler, task processor and state-transition manager, with unit tests.
  • Separated metric-based evaluation from execution, keeping human approval before changes.
  • Related capacity work included alert-triggered scaling callbacks and batch scaling APIs.

State machines · API development · Task scheduling

Coles Group

Retail APIs & transaction services

Coles Group · Team project

Merchandise Domain API

Worked on backend APIs serving product information to retail websites and apps. The service combined downstream domains, using Java virtual threads to parallelize requests.

My work also included infrastructure setup and delivery automation.

Java · Spring Boot · Redis · Terraform · Azure · Kubernetes

Retail website

Coles Group · Team project

Gift card transaction processing

Worked on services that receive and enrich transactions from third-party vendors, then persist them to relational and NoSQL stores.

The implementation used Spring WebFlux, messaging and infrastructure as code, with a Next.js interface for transaction monitoring.

Java · Spring WebFlux · Service Bus · Cosmos DB · Next.js

Research

Graph algorithms & efficient routing

Research · Australian Database Conference 2022

Fuel-efficient navigation

My minor-thesis research explored routing algorithms for finding fuel-efficient paths. I built experiments around graph representations of real-world road data.

The work connects data structures and algorithms with practical transport problems.

C++ · Graph algorithms · Eco-routing · Experimental research

Read the publication

Get in touch

Let's build with AI.

AI applications, agent tools, backend systems, or an interesting software problem to solve.

shain.jobseeking@gmail.com