SHAO-HSIANG
CHIENLet’s talk

AI systems & software engineering

Shao-Hsiang
Chien.

Building AI systems
for real-world work.

System Software Engineer at HP.
Agentic PCs, local inference, and performance.

Explore my work A little about me
Shao-Hsiang Chien / Houston, TXScroll to discover 29.7604° N   95.3698° W
Industry & researchRice UniversityhpMicrosoftSYSTEX

01 Selected work

Independent products, tools & research

Ideas into
working systems.

Useful AI products.
Thoughtful tools for the people building them.

Roomfile’s design example: a warm living room with a sofa, patterned rug, and workspace.

AI Product / 2026

Roomfile

From room inspiration to a practical plan

An interior-design agent skill that remembers your room, develops ideas with you, checks furniture fit, and turns a concept into a sourcing and execution plan.

Agent SkillsJavaScriptSpatial ValidationGenerative Images
View on GitHub
Explore the project

Independent open-source project · v0.3.0

  • Keeps room facts, preferences, and design decisions together across an evolving project.
  • Checks measured furniture footprints against room boundaries, overlap, and clearance zones separately from generated images.
  • Combines researched style guides, visual examples, and product sourcing with a recorded same-room benchmark.
HelmMate project setup screen showing launch state, setup checks, and repository configuration.

Agent Engineering / 2026

HelmMate

A local console for AI coding work

Turns product asks into engineering tickets and coordinates local coding agents through dependencies, isolated worktrees, run tracking, and human review.

ReactNode.jsAgent CLIsGit Worktrees
View on GitHub
Explore the project

Independent open-source project · v0.1.0

  • Carries acceptance criteria, repository context, and dependencies into each agent-ready ticket.
  • Checks launch prerequisites and work-in-progress limits before dispatching local agent runs.
  • Supports Claude Code, Codex, and opencode, with logs and a pull-request handoff for human review.

Developer Tools / 2026

Cairn

Project memory that survives the next session

A lightweight agent skill that interviews a repository, creates a tailored memory layer, and proposes updates as decisions, workstreams, and open questions evolve.

Claude CodeAgent SkillsMarkdownProject Memory
View on GitHub
Explore the project

Independent open-source project · v0 / dogfooding

  • Uses the project’s structure and an interview to choose relevant memory files instead of imposing a fixed template.
  • Preserves existing project instructions while adding a managed memory index and maintenance rules.
  • Proposes a small set of session-end updates for review, with no runtime dependencies.

AI Research / Feb 2025 — Dec 2025

MINT

Causal Tracing of Multimodal Fusion

A causal hidden-state patching framework to localize where visual and textual signals fuse inside the language decoders of Vision-Language Models.

PyTorchVision-Language ModelsLLAVA-1.5DeepSeek-VL2Qwen2-VLLoRA
View on GitHub
Explore the project

ACL 2026 Submission · Rice University Chili Lab

  • Proposed MINT — a causal hidden-state patching framework for localizing multimodal fusion inside VLM language decoders
  • Applied to LLAVA-1.5, DeepSeek-VL2, Qwen2-VL, and Qwen2.5-VL — revealing model-specific fusion bands
  • 100% hallucination correction via fusion-band-targeted LoRA fine-tuning with <50% of standard baseline parameters

02 The person behind the work

Curiosity, with a purpose

Understand
deeply.
Build thoughtfully.

Shao-Hsiang Chien

I’m a System Software Engineer at HP, working on bringing agentic AI capabilities to PCs.

My work spans local inference, performance evaluation, and the practical tradeoffs of running AI on-device. I build the tools and experiments that help turn those tradeoffs into engineering decisions.

Outside work, I build products and developer tools that make AI useful in everyday workflows. My background includes multimodal AI research at Rice University’s Chili Lab, enterprise AI at Systex, and cloud and MLOps engineering at Microsoft.

Education

Rice University

M.S. Computer Science 3.87 / 4.0 GPA

National Tsing Hua University

B.B.A. Business Administration 3.79 / 4.3 GPA

Double specialty in Quantitative Finance & Computer Science

Get to know me
Edge AI✳Agentic Systems✳Inference & Performance✳Full-Stack Engineering

03 Experience

From experiments to engineering decisions.

Built alongside
exceptional teams.

From cloud infrastructure
to adaptive AI systems.

01

2026 — Present

HP Inc.

Focus: Edge AI & Agentic Systems

System Software Engineer

Building agentic capabilities for HP PCs, with a focus on local model serving, NPU-backed inference, and the performance and resource tradeoffs of on-device AI.

Contributions
  • Built a benchmarking pipeline and dataset to compare knowledge-worker agent workloads across models, chipsets, and agent harnesses.
  • Contributed to local inference and model-serving work, evaluating inference performance and system resource use.
  • Supplied measurements from experiments to support marketing’s agent ROI calculator.
Edge AINPU InferenceLocal Model ServingAgent BenchmarkingPerformance Analysis
02

May 2025 — Dec 2025

HP Inc.

Agentic AI Systems — Security R&D

Software Engineering Intern

Built agentic AI frameworks for enterprise security — modular systems that dynamically adapt through intelligent agent architectures and MCP server orchestration.

Contributions
  • Architected an agentic AI framework that dynamically discovers, installs, and orchestrates MCP servers from a cloud registry at runtime — enabling adaptive multi-step workflows without redeployment
  • Designed supervisor-based multi-agent architecture, refactoring a monolithic system into modular, single-responsibility agents to improve scalability, maintainability, and fault isolation
  • Led an approved defensive publication: 'Prompt-Aware Dynamic MCP Loading' — LLM-based intent-aware server loading that minimizes runtime overhead
Multi-Agent SystemsLLMsMCP ProtocolPythonAI Architecture
03

Jan 2024 — Jul 2024

Systex Software

Enterprise AI Platform

Software Engineer

Led development of enterprise-grade generative AI SaaS platform, implementing production RAG systems for Taiwan's largest financial institutions.

Contributions
  • Led full-stack generative AI SaaS with React.js + FastAPI — resulted in potential sale to Taiwan's largest bank
  • Implemented production-level RAG with Azure AI Search and Azure OpenAI through extensive research and optimization
  • Reduced deployment time by 80% via infrastructure as code (Bicep) + Docker containerization
  • Developed Pytest unit tests in agile CI/CD pipeline (Jenkins + Azure DevOps), yielding 40% faster dev cycles
React.jsFastAPIAzure OpenAIDockerBicepJenkins
04

Jul 2022 — Jun 2023

Microsoft

Azure Data & AI

Software Intern

Delivered MLOps workshops to enterprise clients and built CI/CD pipelines for Azure ML, creating cloud demos showcasing Azure Data & AI products.

Contributions
  • Delivered MLOps workshops to 150+ enterprise clients — built CI/CD pipelines for Azure ML, AKS deployment, and monitoring
  • Created 6 cloud demos showcasing Azure Data & AI products, including OpenAI model integrations
Azure MLAzure DevOpsMLOpsAKSOpenAI

04 A conversation starts here

Houston, Texas · Open to collaboration

Have something
in mind?

Practical AI, open-source tools, or an interesting idea.
I’d love to hear what you’re working on.

Start a conversation