AI-Powered Supply Chain

HPE

My Role

Lead Product Designer (0→1 Initiative)

Scope: End-to-end product definition, system design, and AI-driven automation strategy

Timeline 6 months

AI-Powered Supply Chain Platform for Ethical Gold Sourcing

I led the design of a 0→1 AI-powered platform to prevent illegally sourced gold from entering the global supply chain.

The initiative evolved from a transactional tracking tool into a scalable provenance platform that automated authentication using AI visual recognition and enabled multi-stakeholder coordination across miners, logistics operators, and refiners.

Strategic Context

Illegal gold mining contributes to environmental destruction, human rights violations, and global compliance risk.

Problem Statement

How might we design a scalable, trusted digital platform that enables secure provenance verification across distributed stakeholders with minimal friction?

Design Process

Defining the Platform Vision

Rather than designing a single-use app, I reframed the initiative as a multi-actor platform ecosystem supporting:

Source registration (miners)
Transit validation (logistics operators)
Refinement authentication (refiners)
Compliance traceability (regulators, enterprise buyers)This shift elevated the product from a workflow tool to a platformized transaction layer for ethical sourcing.

Multi-Sided User Architecture

Through facilitated workshops with PM, engineering, and compliance leaders, I led the creation of a structured persona framework grounded in:

Context: The environment in which each persona interacts with the app.
Motivation: The specific goals and needs of each persona.
Jobs to be Done: Key tasks each persona must accomplish within the app.
Challenges: Obstacles faced by each persona in achieving their goals.

Through our workshops, we established four criteria for persona development:

Miners

Require simple, low-friction capture flows

Operate in limited connectivity environments

Logistics Operators

Validate custody transfers

Track shipments across regions

Require bulk processing efficiency

Refiners

Authenticate origin

Validate documentation

Ensure regulatory compliance

Storytelling

To ensure cross-functional understanding, I created storyboards and process flow based on contextual inquiries. I mapped the user journey to highlight pain points and opportunities for our solution to alleviate them. This method allowed stakeholders to step into users’ shoes and empathize with their experiences, aligning the team around a cohesive conceptual model before moving to wireframe.

Phase 1: Minimal Viable Experience (MVE)

With the help of the product owner, developers, and QA engineers, we brainstormed to determine usability goals and define the main elements and actions the application should provide for the first release.

Defining Usability Goals

Defining Objects and Actions

Ideation

I created low-fidelity wireframes to validate assumptions with my team.

First Iteration

The initial version relied on RFID tagging for identification.

While functional, it exposed systemic issues:

Manual labor at scale
Seal tampering risk
Repetitive low-value tasks
Limited differentiation

This revealed a core insight:

The bottleneck was not tracking — it was authentication friction.

Phase 2: AI-Driven Automation (0→1 Pivot)

I identified an opportunity to replace physical seals with AI visual recognition, enabling:

Unique fingerprinting of gold bars
Image-based registration
Automated verification
Reduced tampering risk
Scalable authentication

This pivot transformed the product from a workflow-optimization solution to an AI-powered identity infrastructure.

AI Product Strategy

I led structured ideation sessions to:

Identify repetitive manual tasks suitable for automation
Map human + AI collaboration touchpoints
Define explainability requirements
Establish trust feedback loops

Workflow to identify tasks to automate with AI

Example of Brainstorming Ideas Through Sketching

Proof of concept

After brainstorming, I created a proof of concept for the developers and the product manager. I did this at the end of my time in Quenta, and I don’t have any information on the next stage of this initiative.

Proof of concept for AI visual recognition app

Key Design Considerations

AI feedback loop: Instead of designing a static interface, I architected a dynamic feedback loop that leverages user corrections to fuel platform growth.

Human Override Workflows

I designed override flows that allowed users to:

Reject a match
Correct discrepancies

Confidence Scoring Transparency

When the AI visually recognized a gold bar, it generated a probability score indicating the model’s confidence in the match.

Instead of treating AI as a black box, I designed the interface to:

- Display a clear confidence percentage (e.g., 98% match)
- Provide contextual explanations (e.g., Matched based on surface markings and edge patterns.)
- Use visual indicators (color states/thresholds)

Confident score for AI visual recognition

Final Solution

Light theme

Images of an app to track precious metals

Dark theme

Prototype

Design Impact

The first version of the application was launched in 2019, and a handful of miners started using it as early adopters. Then, I worked on a beta version that uses AI visual recognition. I left the company before the AI version was launched.

See the product website

Challenges Faced

For this project, I had to

Aligned engineering feasibility with product vision
Advocated for platform scalability over short-term feature wins

Also, to elevate my design role from execution to strategic contributor, I am proactively learning the fundamentals of two technologies:

Blockchain
AI visual recognition

This enabled me to participate credibly in architectural discussions.