I architected an end-to-end partner management solution designed to seamlessly onboard vendors and proactively manage their performance through automated health tracking for a premium E-commerce platform. This system automates the entire lifecycle—from initial onboarding to AI-driven performance optimization—ensuring high service standards across the vendor network.
"As a Partner Manager, I want a guided onboarding wizard so that I can collect all necessary vendor documents (Legal Name, Estimated Annual Revenue, Business Type) without manual back-and-forth emails."
"As an Executive, I want to see real-time onboarding metrics so that I can identify which type of business in Vietnam are growing the fastest and where we have bottlenecks."
A Reactive Screen Flow standardizes new partner entry, simultaneously creating an Account and Partner Review record via Master-Detail relationship. A Health Score field (1–100) aggregates Speed, Quality, and Responsiveness KPIs into a live vendor performance snapshot.
Einstein Prompt Builder analyzes specific metric deficits per vendor and generates bespoke coaching plans — including vendors seeking a higher service tier. A Human-in-the-Loop interface lets managers review and finalize AI suggestions before auto-dispatch via email.
I translated complex relational datasets into a Service Quality Command Center to drive strategic decision-making:
Partner Review records can be manually rated and stored by the Quality Audit Manager, or a Scheduled Flow can be configured to automate weekly performance tracking, ensuring consistent oversight without manual reminders or missed review cycles.
Eliminated manual back-and-forth with a guided wizard that collects and validates all required vendor documents in a single, standardized flow.
Underperforming vendors receive bespoke, AI-generated coaching plans reviewed by managers — replacing guesswork with structured, data-driven guidance.
Real-time dashboards surface which business types are thriving or struggling, giving leadership a live pulse on the partner network without manual reporting.
This project taught me how to think beyond individual features and design a connected system — where a Screen Flow feeds data into a Health Score, which triggers an AI coaching plan, which is then validated by a human before reaching the vendor. Each component serves the next.
Using Prompt Builder to generate coaching plans pushed me to think carefully about how context is passed to an LLM and how to structure prompts that return consistent, useful output. The Human-in-the-Loop pattern was a deliberate choice to ensure AI suggestions are always verified, a principle I'll carry into future builds.







