CHULIYING

Frontend Lead / Agent Systems Engineer

Chu Liying
emailgithub

ABOUT

AI AGENTS & DEVELOPER TOOLING

Alongside my frontend leadership role at HDRenewables, I lead the team's AI-assisted SDLC and developer tooling, and co-develop a Rust agent runtime. Across these projects, I work on runtime orchestration, tool execution, retrieval, evaluation, and release controls in Rust, Python, and TypeScript.

My recent work includes a Rust agent runtime, a governed RAG product, and skill-commons, an open-source workflow framework for coding agents. I design failure handling, auditability, and human approval into the system, then verify them through behavioral, replay, adversarial, and end-to-end evaluation.

CAPABILITIES

Agent systems
Agent runtime orchestration for MCP tool execution, session memory, structured state, streaming delivery, guardrails, audit, and OpenAI-compatible interfaces.
Retrieval & grounding
Python and TypeScript pipelines, PostgreSQL/pgvector, hybrid RRF retrieval, evidence validation, citation control, and deterministic fallback.
Evaluation & safety
Behavioral, replay, and adversarial tests across model responses and end-to-end application paths, with fail-closed privacy boundaries, regression gates, and CI evidence.
Developer tooling & adoption
Repository-native workflows for requirements, implementation, review, verification, and release across coding agents. Led adoption with engineers, PM, UI/UX, and QA across a roughly 25-person product team, while preserving review evidence and human release control.

EXPERIENCE

HDRenewablesR&D Assistant Manager / Frontend Lead2024–Present
  • -Lead frontend architecture, delivery, and line management for multi-module energy products built with React, Next.js, and TypeScript
  • -Define API, client-state, and streaming boundaries for responsive dashboards and real-time operational interfaces
  • -Guide planning, task allocation, code review, and production debugging across product, design, backend, and QA
  • -Lead the AI-assisted SDLC for a roughly 25-person product team, keeping specifications, prototypes, implementation, and test evidence aligned across roles
EcosTekSenior Frontend Engineer2023–2024
  • -Coordinated frontend delivery through sprint planning, task allocation, and code reviews
  • -Maintained customer-facing storefronts and admin consoles used by around 100 client accounts, integrating payment, logistics, and inventory systems
  • -Built subscription, promotion, and preorder systems, a Shopee migration dashboard, and Cypress automation for promotion QA
Brewing ArtsFrontend Engineer2021–2023
  • -Owned product design and frontend development from concept through launch
  • -Built the platform with Next.js and GraphQL, including static generation and SEO
Independent ProjectsFreelance Product Designer / Frontend Engineer2020–2021
  • -Delivered client projects across product design and frontend development
The News LensFrontend Engineer2015–2020
  • -Worked across product planning, UX design, and frontend engineering for a high-traffic bilingual media platform
  • -Built branded editorial sites and native advertising experiences
  • -Took new brand sites from early product strategy and visual design through frontend implementation

EDUCATION

Ming Chuan University
Commercial Design · 2012

AI ENGINEERING PROJECTS

skill-commons

Created an open-source developer tooling framework for Claude Code, Codex, and Cursor that standardizes AI-assisted software development across requirements, planning, implementation, review, verification, and release.

  • Defined repository-level workflow contracts with deterministic checks, explicit approval gates, and versioned execution evidence.
  • Compared Claude Code, Codex, and Cursor against shared workflow contracts to document differences in skill discovery, adapter behavior, and integration requirements.
  • Led adoption with engineers, PM, UI/UX, and QA, turning production requirements into reusable workflows with reviewed code, documentation, and verification evidence.
  • Used hermetic fixtures, journey evaluations, coverage rosters, and proof-of-use checks to catch false-green workflows.

Outcome

Validated workflow contracts, artifact provenance, and proof-of-use behavior through 189 Python tests and an 18-suite canonical verifier.

  • Python
  • Git
  • Claude Code
  • Codex
  • Cursor
  • CI/CD

Rust Agent Runtime

Internal project

Co-developed a Rust agent runtime that orchestrates model responses, tool execution, session state, streaming, guardrails, and application delivery.

  • Normalized input and coordinated intent classification, slot extraction, MCP tool execution, and session memory through a deterministic path.
  • Unified REST/SSE orchestration behind an OpenAI-compatible chat endpoint with usage tracking and history-aware intent.
  • Added structured audit trails, replay evaluation, regression checks, and rollback gates to diagnose agent failures and prevent workflow regressions.

Outcome

Turned agent behavior into a replayable CI contract so orchestration and API changes could be reviewed before release.

  • Rust
  • MCP
  • REST/SSE
  • OpenAI-compatible API
  • GitHub Actions

Governed CRM RAG

Internal project

Designed and shipped a retrieval product that answers from approved aggregate data while keeping customer-level data away from external models.

  • Connected Python and TypeScript ingestion to PostgreSQL/pgvector, BGE-M3 embeddings, and hybrid RRF retrieval.
  • Checked citations and evidence, enforced structured output, and returned deterministic fallbacks for unsafe or invalid responses.
  • Kept provider calls behind least-privilege database functions, evidence-bound authorization, and fail-closed privacy gates.

Outcome

Enabled retrieval over approved aggregate data while keeping customer-level data outside external-model calls, with authorization, citation, fallback, and privacy checks in the release path.

  • Python
  • TypeScript
  • PostgreSQL/pgvector
  • Cloudflare Workers AI
  • RAG