AI product / solution engineer
Peter Liu
AI Product /Solution Engineer
My work sits close to real business operations: LLM applications, agent workflows, and local tool delivery. Recent projects include a public-account content workbench, an agent operations console, and a voice visitor-registration flow. This page lists what I built, the stack I used, and what was delivered.
Content workbench, business-process automation, voice agent
James Cook University Singapore, International Business
Native Chinese; English can be used as a working language
Projects
Project Experience
Only public examples are listed here. The GitHub repositories are sanitized versions and do not include client data, API keys, runtime logs, or historical generated content.
ZephyrPress Forge / Wind-industry public-account content workbench
Local-first public-account weekly report workflow
Built for wind-industry content operations. Public information collection, material screening, human review, template-based drafting, and public-account HTML export are kept inside a local workflow. The final publishing decision stays with the operator.
ClawPress Ops Console / Agent content workflow console
Public-account content operations console
Moves an agent-driven content pipeline out of scripts, folders, and chat history into a visible operations console. Operators can review task queues, content records, search strategies, diagnostics, and generated HTML files.
BlueWhale Voice Agent
Voice registration system for park visitors and vehicles
Built for an industrial-park gate scenario. When a visitor vehicle arrives, the caller speaks with a voice agent that captures the plate number, visiting company, phone number, and visit reason. The backend creates a registration record and sends it to the guard for confirmation.
Scope
What I Usually Handle
In actual projects, my work usually covers these parts.
Requirements and process breakdown. Clarify users, input materials, confirmation points, failure handling, and final deliverables before deciding the page and data structure.
LLM / agent application delivery. Connect model output to pages, task status, tool calls, and data structures, so results can be reviewed, edited, and passed into the next step.
Local tool delivery. Delivered local HTTP services, SQLite data storage, Windows install scripts, ZIP installers, upgrade packages, and acceptance scripts.
Documentation and handoff. Prepare user guides, installation notes, acceptance paths, and follow-up improvement lists, so business users can keep using and evaluating the tool.
Background
Basic Information
Position
AI product / solution engineer, focused on LLM applications, agent workflows, local tools, content automation, and business-system delivery.
Stack
React, Vite, TypeScript, Node.js, Python, FastAPI, SQLite, LLM APIs, OpenAI-compatible APIs.
Education
James Cook University Singapore, International Business. In technical projects, I pay close attention to business process, users, and whether the final tool can be handed over for continued use.
Languages
Native Chinese. English can be used as a working language for documentation search, product notes, and basic cross-team communication.
Contact
Contact
For the full resume or project materials, email is the best way to reach me.