Melbourne, Australia

Cheng Yu.

Senior Software Engineer / Tech Lead - AI Systems

I build reliable platforms, AI agents and the tools that connect them.

Tech Lead at Unitix, working on Merchi, and developing Cestiny AI part-time. 10+ years building business platforms and APIs with Python/Flask and hands-on TypeScript/Node.js development. Recent work includes configurable LangChain and MCP agents, business automation and code-repair pipelines with human approval and automated tests.

Download resume LinkedIn
Cheng Yu

Experience

Cestiny AI

AI Product Founder & Lead Engineer (Part-time)AUG 2026 — PRESENT

Leading AI product engineering for Cestiny AI on a part-time basis alongside my main role, building a BaZi analysis and chat tool informed by in-person consultations.

Timeline

  1. Aug 2026 — Present

    Cestiny AI · Part-time venture

    Building an early-stage BaZi software venture alongside my main role, combining AI-assisted programming with domain knowledge.

    • Developing a rules-based harness that constrains AI interpretation within BaZi domain knowledge, supporting an analysis and chat tool.
    • Built unit tests for domain-rule behavior and edge cases, and refined structured evidence queries for AI interpretation.
  2. Mar 2026 — Present

    In-person BaZi consultations

    Began offering in-person BaZi consultations as a side practice, which informed the direction of the software venture.

More milestones and project details will be added over time.

Unitix

Tech LeadAUG 2019 — PRESENT

Leading development of Merchi, a commerce platform connecting sellers and suppliers of customizable merchandise across multiple storefronts and domains, with APIs, SDKs and MCP integrations.

  • Developing an AI Tester and building Programming Agent workflows for automated testing, code repair and human-reviewed PRs.
  • Built configurable customer and manager bots with LangChain and MCP, shared API permissions and approval controls.
  • Developed TypeScript SDKs and Node.js tooling for WebSocket integration, component compilation and release automation.
  • Introduced two-week Scrum iterations after returning as Tech Lead in 2019, incorporating user feedback into development planning; introduced test-driven development.
  • Reduced CI/CD pipeline time by 90%; optimized container builds and caching to improve delivery efficiency and lower CI/CD running costs.

Timeline

  1. Sep 2026 — Present

    AI Tester · In development

    View project details

    Developing an AI tester that automatically tests the project and provides actionable feedback, designed to pair with the Programming Agent in a continuous improvement loop.

    • Working toward a continuous cycle of testing, code repair, and revalidation so the two agents can improve the codebase together.
  2. Jun — Sep 2026

    Programming Agent

    View project details

    Built a Programming Agent harness and tool pipeline that delegates errors and GitHub Issues to Cursor cloud agents, producing tested fix PRs for human approval.

    • Monitored backend errors, investigated the relevant code, and automatically opened fix pull requests.
    • Read GitHub Issues and implemented fixes for issues the agent could resolve, running tests as part of the development loop.
    • Asked the team for clarification when an issue required additional input, shifting day-to-day work toward managing GitHub Issues and providing feedback.
    • Implemented operator-approval checks for issue closure and existing PR merges, with unit tests for approval and queue-state behavior.
  3. Nov 2025 — Jun 2026

    Manager Bot

    View project details

    Built a configurable job-management agent for supplier coordination, shipment creation and scheduling, with manager notifications and approval for significant actions. Manager feedback indicated approximately 30% improvement in productivity.

    • Labeled cases needing human input and notified store managers; significant actions require manager approval.
    • Made the level of automation configurable per store, including disabling agent activity.
    • Manager feedback indicated approximately 30% improvement in productivity; this is an estimate, not a formal time study.

    Technologies: MCP · mcp-use · LangChain

  4. Mar — Nov 2025

    Client Bot

    View project details

    Built a customer chatbot for Merchi's multi-site commerce platform using mcp-use, LangChain and MCP, with store-specific behavior configured through owner-provided skills files. At Custom Wristband (customwristband.com.au), AI now handles over 50% of chats, with some customers placing orders through the bot.

    • Connected MCP tool calls to authenticated user context and the API's shared permission system, backed by existing API permission regression tests.
    • Capped customer-facing bot access at client-level permissions while preserving more restrictive user permissions.
    • Integrated LangChain chat models and message history with mcp-use agents, contextual instructions, tool restrictions and token-usage callbacks.
    • Enabled store owners to provide skills files that customize their site's client bot, tailoring customer interactions across the platform's stores and domains.

    Technologies: MCP · mcp-use · LangChain

  5. 2025

    MCP support

    Added Model Context Protocol (MCP) support as part of integrating AI with the platform.

  6. Early 2025

    Code optimization and security refactoring

    Optimized the existing codebase and refactored legacy code to address security issues, preparing for further AI integration.

  7. Early 2023 — end of 2024

    Site Builder

    View project details

    Built the Python orchestration and Node.js compilation pipeline for a storefront editor using database-stored React components, with frontend syntax validation and compilation feedback.

    • Implemented frontend syntax validation and surfaced compilation errors in the browser editor; a colleague designed the browser editor interface.
    • Supported customer-managed deployment or hosting on our platform.

    Technologies: React.js

  8. Mar — May 2021

    AWS and CI/CD optimization

    Reduced AWS costs by 80% over approximately two months while also shortening CI/CD pipeline time and lowering its running costs.

    • Optimized container builds and caching to make CI/CD more efficient.
  9. Early 2020 — Feb 2023

    Value Chain System

    View project details

    Developed a supply-chain platform enabling influencers and sales partners to select products to sell, with direct fulfillment by the platform and revenue sharing under agreed terms.

    • Developed workflows for influencers and sales partners to select products, sell to customers, and have the platform fulfill and ship orders directly.
    • Worked on a tiered supply-chain structure and revenue-sharing arrangements between participating parties.
  10. First six months after returning

    CI/CD and React architecture

    Introduced continuous integration and continuous delivery (CI/CD) into the project’s development and release workflow.

    • Introduced Redux and Redux-Saga into the existing React architecture during the same period.
  11. Aug 2019

    Returned as Tech Lead

    Returned to Unitix following the previous technical lead’s departure, taking over technical leadership and supporting the Merchi platform.

More milestones and project details will be added over time.

Explore Merchi ↗Python / PHP SDK — https://github.com/merchisdk/sdk ↗JavaScript SDK — https://github.com/merchisdk/merchi_sdk_js ↗TypeScript SDK — https://github.com/merchisdk/merchi_sdk_ts ↗
Full Stack DeveloperAUG 2014 — OCT 2017

Developed Merchi’s Flask and PostgreSQL API framework and core business functionality through its 2016 launch, followed by warehouse and logistics features and system containerization work.

Timeline

  1. Final months — Oct 2017

    System containerization

    Advanced the project’s containerization design and implementation alongside business bug fixes. The migration was still in progress when I left Unitix in October 2017.

  2. First half of 2017

    Warehouse and logistics

    Developed warehouse and logistics functionality during the first six months following launch.

  3. After May — late 2016

    Core business development and launch

    Completed the main business functionality built on the framework and helped launch merchi.co by the end of 2016.

  4. Aug 2014 — May 2016

    API framework and SDK development

    Focused primarily on developing and maturing Merchi’s custom Flask and PostgreSQL API framework.

    • Enabled frontend clients to request backend resources on demand—a concept similar to GraphQL’s approach—with reusable CRUD operations and permission management.
    • Led JavaScript and Python SDK development to simplify API integration.

More milestones and project details will be added over time.

Python / PHP SDK — https://github.com/merchisdk/sdk ↗JavaScript SDK — https://github.com/merchisdk/merchi_sdk_js ↗

Common Code

Software DeveloperOCT 2017 — JUL 2019

Worked across client projects with Django, React, and React Native, delivering agriculture, real-time application, and waste-management software.

Timeline

  1. May 2019 — Jul 2019

    Dimeo

    View project details

    Worked on a waste-management system using Raspberry Pi devices installed in bins, with Django-based management and resin.io (now balenaCloud) for fleet management.

    • Supported a collection workflow for photographing waste, measuring its weight, and centrally managing collection records.
    • Troubleshot tricky device-level issues, including inaccurate measurement readings, and made hardware-related adjustments and optimizations.
    • Worked with a fleet of more than 40 Raspberry Pi devices in a system supporting over 2,000 users.

    Technologies: Django · Raspberry Pi · resin.io / balenaCloud

    Company website ↗

  2. Jan 2018 — Apr 2019

    Agrinous

    View project details

    Core contributor during Agrinous’s early development, primarily responsible for the backend of an offline-capable livestock application.

    • Designed and developed Django REST APIs supporting a React Native application.
    • Improved performance of optimized API endpoints by approximately 100–1,000×.
    • Supported cattle and sheep counting in rural locations with unreliable or unavailable internet access.
    • Worked on the backend supporting offline field records and synchronization to the cloud when connectivity returned.

    Technologies: React Native · Django REST API

    Website ↗

  3. Oct 2017 — Dec 2017

    MEERQAT

    View project details

    Maintained MEERQAT, built with Meteor.js, React.js, and Django DDP, addressing integration and upgrade-compatibility issues caused by legacy dependencies.

    Technologies: Meteor.js · React.js · Django DDP

    Website ↗

More milestones and project details will be added over time.

Dimeo ↗Agrinous ↗MEERQAT ↗Common Code ↗

Skills & education

AI engineering

  • Model Context Protocol (MCP)
  • Configurable agent skills
  • LangChain / mcp-use
  • AI agents / tool integration
  • Domain-rule harnesses
  • AI testing workflows

Software engineering

  • Python
  • Flask
  • SQLAlchemy
  • SQL
  • Django
  • PostgreSQL
  • React
  • AWS / GCP
  • Docker
  • Motoko / ICP
  • TypeScript
  • Unit testing / TDD

Master’s degree, Computer Science

RMIT · 2012 — 2014

GPA 3.8

Research & scholarships

  • Sentiment analysis research & exhibition

    Research completed March 2014

    A$4,000 research exhibition funding

    Completed further research into machine learning for sentiment analysis and labeling under the supervision of Xiuzhen Jenny Zhang in March 2014, before joining Unitix that August. Presented the research at RMIT’s university–industry research exhibition with A$4,000 in funding.

  • AI-assisted data labeling

    December 2013

    Summer Internship · A$4,000 scholarship

    Researched active labeling under the supervision of Xiuzhen Jenny Zhang, exploring how machine learning could assist data annotation and reduce manual labeling effort.

  • University–industry collaboration

    First half of 2013

    A$10,000 scholarship

    Led a collaborative project between RMIT and Lexer under the supervision of Xiuzhen Jenny Zhang.

    Industry partner website ↗

  • Search engine index optimization

    December 2012

    Summer Internship · A$4,000 scholarship

    Completed a research internship investigating search engine index optimization under the supervision of J. Shane Culpepper.

Bachelor’s degree, Computer Science

Hubei University of Technology · 2007 — 2011

GPA 3.2

Open source

mydocker · 2018

2,000+ GitHub stars

Contributed three merged fixes and improvements to an educational Go container engine, covering namespace-related imports, networking API compatibility and workspace cleanup.

View contributions

Get in touch