刘浩然Product-Focused Tech Lead

From business judgment to product delivery.Technology that solves real business problems.

AI / AGENTProduct direction

Focused on AI and Agent products: bringing AI from business understanding into decisions and action, with control and reliability built in.

View selected casesLet’s discuss your product
Explore how I work

02 /APPROACH

Super Individual → Super Team

Connect product judgment, critical implementation and collaborative delivery.

My contribution

AI contribution

Team collaboration

  1. Which problem is worth solving?

    • My contributionIdentify the problem worth solving
    • AI contributionOrganize material and support analysis
    • Team collaborationClarify the scenario with the business side

  2. What to build, and where to start?

  3. How to build it and implement what matters?

  4. How to validate, deliver and evolve?

Choose independent development or team collaboration to suit the product stage, work directly on architecture and critical implementation, use AI to support engineering feedback, and coordinate specialist delivery.

03 /SELECTED WORK

Selected cases

AI products, complete applications and complex systems in practice.

Natural-language control for smart hardware

Make natural language a control interface for devices

Natural expressions do not map directly to device operations. The same instruction can need different handling depending on device capabilities and operating state.

What the user says

Intent · capability · state

Structured command, or clarification

SchematicUtterance ▸ intent · capability · state ▸ command / Utterance ▸ unclear intent ▸ clarify / Utterance ▸ unsupported ▸ feedback

Key judgment

Language understanding is never treated as a device action: every command stays within device capability and current state.

Delivered capabilities

Delivered an app-facing core AI service that translates natural language into structured control commands, completed integration and passed acceptance.

Explore case

Approach and implementation

Combined intent recognition with device capabilities and current state to generate commands within operating constraints, with clarification and feedback for ambiguous or unsupported requests.

Contribution

Core AI service / Independent design and development

Equipment repair knowledge assistant

Make repair answers traceable to their sources

Repair questions draw on manuals, fault information and practical knowledge, with technical terms and fault codes requiring precise matching.

Manuals · fault information · field knowledge

Exact + semantic retrieval

Answer linked to its sources

SchematicQuestion ▸ terms · fault codes ▸ answer + sources / Question ▸ insufficient material ▸ ask to confirm / Question ▸ unclear scope ▸ answer boundary

Key judgment

Every answer must trace back to a source; when material is thin or scope is unclear, the assistant confirms first instead of guessing.

Delivered capabilities

Organized repair manuals, fault information and practical knowledge into a Q&A service, linking answers to source material so users can find technical information and check its basis.

Explore case

Approach and implementation

Combined exact and semantic retrieval to organize relevant material and link answers to sources, with clarification and answer boundaries when information or applicability is unclear.

Contribution

Knowledge retrieval / Answer grounding and boundaries

Health services across user and store applications

Connect user services, store operations and commerce

As the business grew, the original application structure struggled to take on new product requirements, and local patching could no longer support continued expansion.

User app · store app

One product architecture

Cloud services: health and commerce

Key judgment

Judged that a rebuild would serve better than further patching, and connected every client through one architecture.

Delivered capabilities

Delivered user and store applications with cloud services connecting health services and commerce, completing the platform rebuild and subsequent product iterations.

Explore case

Approach and implementation

Judged that a full rebuild would serve better than further patching, and proposed and led it: redefined business boundaries around user services, store operations and commerce, connected every client through one product architecture, and used AI Coding for implementation, review and iteration.

Contribution

Architecture / Independent full-stack development

Financial asset management platform

Trace transactions and maintain consistent asset records

Fund flows, subscriptions, holdings and redemptions span multiple states and external services, requiring review, execution and records to stay connected.

Fund flows · external notifications

Review → execute → record

Traceable transactions, consistent records

Key judgment

Duplicate-notification handling, amount precision and transaction state are core design, not afterthoughts.

Delivered capabilities

Delivered a transaction backend covering fund flows, subscriptions, holdings and redemptions, integrated third-party custody services, and supported transaction tracing and consistency checks across fund and asset records.

Explore case

Approach and implementation

Validated notification sources and prevented duplicate processing, combining monetary precision and transaction state management to connect third-party custody services with internal asset records.

Contribution

Transaction backend / Custody integration and data consistency

More project experience

Consumer device software and content platform

Connect device interaction, content adaptation and sharing

Device integration and content adaptation

Explore case
The problem

Content must meet different device requirements, while software interactions need to reflect device state and responses.

Approach and implementation

Adapted content to device requirements, included state and error feedback in interaction design, and connected software actions to device responses through integration testing.

Result

Delivered Windows companion software and a cloud content platform for device interaction, multimedia content adaptation and sharing, refining the software through feedback after releases.

Contribution

Desktop software / Device integration and content platform

Multimodal information and report generation

Turn field material into reviewable reports

Report generation and human review

Explore case
The problem

Research recordings, photos, documents and notes need to be organized into a report structure and checked.

Approach and implementation

Separated transcription, recognition and report writing, organized information around the report structure, and retained human review and additions.

Result

Combined recordings, photos, documents and notes into structured report drafts, with a shared workspace for corrections and export in business report formats.

Contribution

Multimodal processing / Report generation and human review

Medical imaging platform

Support imaging access, cross-institution sharing and workflow monitoring

Data architecture and business-flow monitoring

Explore case
The problem

Growing imaging data and cross-institution access placed increasing demands on processing, scalability and service reliability.

Approach and implementation

Balanced performance, scaling and maintenance costs in the data architecture upgrade, and extended monitoring from server health to business data flows across imaging workflows.

Result

Supported imaging access and cross-institution sharing, upgraded the data architecture, and used business data-flow monitoring and alerts to identify anomalies such as data failing to arrive for an extended period.

Contribution

Engineering leadership / Architecture and reliability

Fresh-supply multi-tenant SaaS

Connect procurement, orders and picking in one product

Multi-tenant architecture and business isolation

Explore case
The problem

Procurement, orders, warehousing and picking are closely connected. Serving multiple businesses requires accommodating workflow differences, data isolation and growing system load.

Approach and implementation

Translated shared workflows and business differences into a multi-tenant product, using a full rebuild and core architecture design to address data isolation, extensibility and growing system load.

Result

Brought procurement, orders, warehousing and picking into a multi-tenant supply-chain product, supporting connected workflows and data isolation across businesses through a product rebuild.

Contribution

Product and core architecture / Product and engineering delivery

04 /EXPERIENCE

Experience & capabilities

From complex trading systems to enterprise products and engineering organizations, then AI product practice. Experience through business evolution, with hands-on work in architecture and critical implementation.

  1. 2010–2014

    Contributed to and led trading system development, building experience in high concurrency, asynchronous processing and service-oriented architecture.

  2. 2014–2021

    Connected product planning, architecture and team delivery through successive stages of business development.

  3. 2021–2025

    Led medical product engineering, architecture and delivery, advancing reliability practices and compliance coordination.

  4. Since 2025

    Building products independently and using AI Coding in development, with a focus on AI and Agent applications.

    • Since 2025Independent product & engineeringShenzhen Haobing Xingcheng Technology Co., Ltd.Independent product development and collaborative delivery, focused on AI and Agent applications.

AI-assisted product development

Integrate AI into analysis, implementation, review and validation, while owning architecture, key decisions and delivery quality.

AI Coding · Claude Code / Codex

AI applications & controlled execution

Combine model interpretation with business context, device capabilities and state to produce operations with clear boundaries that can be validated.

Model integration · Context handling · Command generation

Complex systems & engineering delivery

Design services, data systems and asynchronous processing around business scale and evolving needs to support product delivery.

Service architecture · Data systems · Multi-platform applications

Selected technologies

  • Java / Spring
  • Python / FastAPI
  • TypeScript
  • React / uni-app
  • Flutter / Tauri
  • MySQL / TiDB
  • Redis / Docker
  • Claude Code / Codex

South China University of TechnologyPMP credential

Enterprise AI Coding training experience: engineering collaboration, quality review and practical development workflows.

05 /COLLABORATE

Build AI productswith business value

Start with the problem you want to solve.

AI / Agent product co-creation

You have a business scenario and want to turn it into a product.

Start with the business problem and product definition, validate key capabilities together, and build and deliver the product.

Reach out this way

Technical partnership

Your product direction is clear and you need ongoing technical involvement.

Collaborate on technical direction, architecture and engineering organization to evolve the product together.

Reach out this way

Product & technical advisory

You face a key decision and need judgment or a solution review.

Get judgment, solution reviews and team support for AI product positioning, technical plans and engineering practices.

Reach out this way

I value collaborations with real business scenarios, clear value goals and a willingness to validate together. Clarify the problem first, then agree on how to collaborate and what to deliver.

Let’s discuss your product

06 /CONTACT

Learn more & get in touch

Contact me directly

Briefly describe the problem, who the product serves and where you would like me to contribute.

patrick.lau@foxmail.com
Send an email

Shenzhen · Remote

Want to learn more?

This website’s AI assistant can explain Patrick’s projects, working approach and collaboration options using the public information on this site.

Website AI assistant · public information only
Website AI assistant · public information only
Website AI assistant

Choose a question or type your own.

ask ·