Devin TsoOperating Portfolio

Business logic / system architecture / AI build

Beyond Complexity,Toward Capacity.

越过纷繁，终抵宏量。

Bridging Financial Rigor with System Architecture.

I turn business complexity into systems that can run, scale, and be trusted.

View ShowcaseContact now

Foundation / real operating evidence

The Logic of Gravity

Before I built systems, I learned where business logic must never fail: transaction flows, settlement rules, reconciliation evidence, workflow ownership, and global delivery standards.

Settlement & risk logicTencent

High-volume channel work trained my sense of auditability, reconciliation, rule clarity, and control points.

Open caseCampaign economicsROI

Activity review connected orders, rebates, subsidies, ARPU, and resource allocation into business decisions.

Open caseCode as precisionPython

Before AI became a shortcut, code was already my way of forcing clarity into noisy reconciliation and multi-source matching.

Open caseGlobal workflow disciplineDHL

Cross-border delivery taught me how to turn requirements, handoffs, and responsibilities into standardized executable workflows.

Open case
Order / delivery flowsVIP consumption ratiosSettlement evidenceRebate / subsidy rulesCross-border workflowsAudit-ready logic

Sanitized code fragments

The early code was not decorative. It turned reconciliation pressure into repeatable precision.

01monthly channel flow
files = glob("channel_flow_*.csv")
flow = pd.concat(
    pd.read_csv(file, encoding="gb18030")
    for file in files
)

pivot = flow.pivot_table(
    values="package_amount",
    index=["product_line", "channel"],
    aggfunc="sum"
)
02missing-order recovery
joined = pd.merge(
    dev_export,
    official_flow,
    how="left",
    left_on="order_id",
    right_on="internal_order_no"
)

unmatched = joined[joined["internal_order_no"].isna()]
03settlement output
with pd.ExcelWriter("settlement_review.xlsx") as writer:
    member_pivot.to_excel(writer, "member_flow", index=False)
    sports_pivot.to_excel(writer, "sports_flow", index=False)
    audit_result.to_excel(writer, "review_result", index=False)

Synthesis / from business reality to running systems

Business Understanding into Running Systems.

I read business through orders, contracts, roles, exceptions, workflows, data, and control points, then translate them into architecture, interfaces, permissions, and working code. Strong delivery also needs active correction: clarifying assumptions, challenging vague logic, and turning feedback into structure.

01
Business understanding

Orders, contracts, roles, exceptions, ownership

02
Logic modeling

Objects, relationships, status, validation, review points

03
System architecture

Interfaces, permissions, workflows, dashboards

04
Code & IT execution

Scripts, pages, tools, integrations, reusable patterns

05
AI-accelerated build

Agents and Codex turn rules into faster iterations

06
Collaborative correction

Frame, test, challenge, revise, and close the loop

b2b-ops.local / operating-console
Business ObjectPO-OPS-0418
In use
CustomerMasked account
SupplierQualified vendor
StatusControlled flow
SKUQtyTaxStatus
VALVE-01120CheckedClear
PUMP-18240CheckedClear
SEAL-09120CheckedClear
Header + line itemsWorkflow statesPermission boundaryBusiness analytics

Current result / already in use

The Masterpiece:Custom B2B Operating System

Not a finance tool, but a full business operating system: customers, suppliers, contracts, orders, fulfillment, settlement, permissions, risks, and analytics connected through structured data. Finance-aware by design, built for the full business flow.

01

A mature system already in motion

The current B2B system is not a demo. It has moved from idea and data cleanup into practical use.

View case
02

Full business objects

Customers, suppliers, products, contracts, orders, fulfillment, settlement, risks, and analysis are connected as one operating layer.

View case
03

Header + Line Items model

Scattered rows become structured records that can support order detail, status, responsibility, and later analytics.

View case
04

Workflow and permission boundaries

The system carries roles, states, handoffs, review points, and operational control instead of only recording results.

View case
05

Finance-aware control points

Settlement, payment terms, tax-related risks, and evidence review remain embedded without narrowing the system to finance.

View case
06

A growing code-built portfolio

This system is the first visible proof of a broader direction: more tools, more software, more business logic turned into code.

View case

Future / AI build loop

AI-Driven Delivery, Human-Centric Logic.

AI accelerates generation. I define the business logic, system boundaries, data structure, trust criteria, and the correction loop that keeps the build aligned with reality.

From one mature B2B system to a growing portfolio of code-built business tools.
01

Define

Clarify the real business object, owner, state, and risk boundary.

02

Generate

Use AI and Codex to turn rules into pages, flows, checks, and tools.

03

Review

Test whether the result matches business logic, user behavior, and control needs.

04

Correct

Challenge vague assumptions, refine wording, adjust scope, and rebuild quickly.

05

Close Loop

Move from prototype to operating rhythm, data feedback, and the next code work.

Contact Now

Let’s BuildWhat Lasts.

Business complexity deserves systems that can hold it.

For system work, business logic, data foundations, and AI-assisted delivery.
Email is released only after a Cloudflare check.