Integration Team - Henry Onboarding Meeting

Source: Notion

Date: 2026-01-29

Attendees: Henry, Tingsong, Monkey

1. Background

Henry is joining the Integration team. This meeting was hosted by Tingsong to introduce the team’s current business status, technical challenges, and architectural transformation direction, while clarifying Henry’s role and initial tasks.

2. Team Core Value

Integration team’s core mission: Connect 30+ third-party Vendors to eliminate manual data entry for financial advisors.

• File-based Vendors: ~18
• API-based Vendors: ~15

3. Current Technical Challenges

3.1 Technical Debt

Architecture decay:

• Legacy PHP code (2015-16), patched for years without systematic refactoring
• Inheritance chains 6-8 levels deep; changes for one Vendor cascade unpredictably to 20+ others
• Over-reliance on static functions; classes lack single responsibility

Black-box debugging:

• Write-time transforms bury logic in object state mutations → intermediate states invisible
• DB_soft_transaction swallows all exceptions, forces loops to continue → cannot break on error
• Each node opaque; tracing a data issue requires hours of code reasoning

3.2 External Vendor Unreliability

• Vendor dev teams (often offshore) have inconsistent engineering quality
• Poor data quality: wrong data types (e.g., card numbers as amounts), unstable payloads
• Breaking changes without notice: sudden type/interface changes crash our system

3.3 Operational Burden

• Data Issues doubled last year despite no new Vendors added
• Team becoming a “cost center” — engineers trapped in manual analysis loops
• Timezone gap with US Support → multi-day round-trips for issue resolution

3.4 Functional Limitations

• No support for negative assets (blocks high-frequency trading clients)
• Legacy rushed projects (e.g., Schwab) still drain maintenance effort years later

---

4. Architecture Transformation

4.1 ETL → ELT

Moving complex transformation logic from PHP to SQL (Snowflake):

• Small T: Trim redundant data and flatten structures before Load
• Large T: Business modeling via SQL Views in Snowflake (optional)

4.2 DITS (Data Issue Tracking System)

Goal: Reduce issue investigation time from hours → minutes

• End-to-end tracing
• Data node transparency
• Enable Support team and even Advisors to self-check data flow

4.3 Tech Stack

• Cloud: AWS
• Data Warehouse: Snowflake

5. 2026 Core Priorities

1. Crypto
2. Schwab
3. Yodlee
4. DITS

6. Expectations for Henry

6.1 Role

Technical Expert: Share the burden of architecture design and deep refactoring with Tingsong.

6.2 Specific Expectations

Area	Content
Advanced perspective	Leverage Data Warehouse, Snowflake, Parquet experience to improve performance
Engineering quality	Introduce DI, single responsibility, Immutable Objects; eliminate static methods
Proactive problem discovery	Review existing systems, identify root causes of slow issue investigation

7. Action Items

Immediate

Action	Owner	Priority
Read Integration 2026 Roadmap	Henry	P0
Address Cloudflare Turnstile project review feedback	Henry	P0
Use Yodlee as entry point to learn the business (continuity with previous SSE work)	Henry	P0

Short-term

Action	Owner	Priority
Research ELT model performance bottlenecks in API-based scenarios	Henry + Monkey	P1
Participate in DITS mechanism design and implementation	Henry	P1
Frequent communication with team to build consensus	Henry	P1

Mid-term

Action	Owner	Priority
Attempt to clean up DB_soft_transaction and other technical debt	Henry + Tingsong	P2
Gradually migrate Vendors to new architecture	Team	P2

8. Why Yodlee as Entry Point

1. Relatively independent business: Account aggregation platform with clear data flow
2. Continuity with SSE work: Henry’s previous experience is transferable
3. 2026 core project: High priority, clear value