Sinch's public role description pointed to a platform problem common in businesses built across multiple product lines: different schemas, APIs, reporting interfaces and metric definitions make cross-product data difficult to understand and use consistently.
I treated internal product teams as data producers and external customers as data consumers, then designed the work around the contracts, trust and alignment needed between both sides. The user archetypes are assumptions created for the exercise, not findings from customer research.
What needed to change
The challenge was to propose one trustworthy layer for dashboards, APIs, exports and later AI while accounting for strict tenant boundaries, different product-team dependencies and the organizational reality of aligning contributors without direct authority.
How I approached it
- Framed the platform mandate, mapped fictional customer archetypes and internal producer stakeholders, and connected their jobs to a shared product strategy.
- Created 15 linked workstreams covering the North Star, market research, vision, value proposition, canonical event model, API principles, access controls, AI architecture, prioritization, roadmap and OKRs.
- Specified a canonical CommunicationEvent, REST and OAuth patterns, cursor pagination, bulk exports, executor-enforced tenant isolation, field masking and audit logging at a product-architecture level.
- Synthesized the work into a 33-page strategy, a 21-slide final deck and a six-screen clickable concept prototype using mocked data.
The choices that shaped the work
Build trust before features
A wrong metric can be worse than a missing one, so consistent definitions, data accuracy and contribution contracts come before expanding the feature surface.
Put the schema and API before dashboards
An immutable event model and stable access contract create a foundation that multiple interfaces can use without redefining the data each time.
Constrain AI through the architecture
The proposed AI flow produces a schema-checked QuerySpec rather than raw SQL, runs through the same permission-aware executor and returns provenance with the answer.
Sequence by evidence and gates
Alerts, warehouse connectors and natural-language querying stay behind explicit readiness gates until definitions, access controls and baseline behavior are dependable.
What the work demonstrates — and what it does not claim
The result is a coherent candidate work sample that connects product strategy, platform architecture, delivery sequencing and an interactive concept; it is not a deployed Sinch product.
The plan makes the organizational dependencies visible by treating data contracts, integration waves and contributor alignment as product work rather than background implementation detail.
The prototype demonstrates the intended experience with mocked data. Its personas, metrics and roadmap figures are assumptions or targets to test, not customer evidence or realized outcomes.
WHAT THIS REINFORCEDFor data platforms, trust is cumulative: consistent definitions, contributor contracts, access controls and provenance have to work before dashboards or AI can create credible value.