I build and scale products in complex, ambiguous environments.

The company

• A trading platform for the jewelry industry. Wholesalers, manufacturers, and designers list diamonds, gemstones, and finished jewelry.
• Retailers source from them and show the goods to their own customers, with rich media and multiple ways to view every item.

The problem

• Retailers want to offer more than they can afford to stock.
• Wholesalers want a convenient way to sell into their retailer networks.
• Both want to sell to their networks and customers online, including goods they do not hold.
• Before VDB this ran on phone calls, spreadsheets, JPG files, and other people's catalogs. Fragmented, slow, impossible to show a customer, awkward to mark up.
• The platforms that existed were ugly, hard to use, or expensive.

Who it is for

The jewelry industry, in these roles:

• Wholesalers
• Retailers
• Manufacturers
• Designers
• Brokers

What I did

• Built the analytics framework from scratch. Instrumentation, metrics and KPIs, user journeys. Used it to find the power users, interviewed them, and tested monetization ideas through design sprints and prototypes.
• Shipped the core trading experience. Search, filtering, saving, sharing, comparing, matching, and buyer-to-seller communications.
• Built the deeper trading workflows. Bidding, multi-currency support, and the full checkout flow.
• Ran the customer lifecycle features. Onboarding, reminders, and everything an e-commerce app needs to keep people coming back.
• Designed the white-label platform with feature switches, so one codebase powers 135+ branded apps and storefronts.

Building the team

• Hired and onboarded a product and design team of seven.
• Set the prioritization framework and the templates for specs, research, and releases.
• Ran the agile process. Planning, reviews, and retrospectives.
• Coached product managers to own their areas end to end.

Impact

Gallery

The VDB marketplace app in use

VDB marketplace for buyers on phone and tablet

Ring Creator walkthrough on a white-label store

White-label storefront on desktop, tablet, and phone

VDB ecosystem: inventory into VDB Cloud, out to B2B apps, white-label tools, and B2C stores

The starting point

• A functional prototype. Two cameras and an edge unit on a forklift, detecting pedestrians and warning the driver.
• A strong hypothesis, unvalidated. The team knew the technology worked. Nobody yet knew who would buy it, or what they would need it to do.
• My brief as Head of Product: validate the idea with real customers and turn the prototype into a product.

Who I listened to

The people responsible for preventing accidents on the warehouse floor:

• EHS managers
• Safety managers
• Operations leaders
• Forklift operators

What they taught me

• EHS managers are stretched thin. They cannot watch footage. Capturing everything is not enough. The system has to find the dangerous moments, show only those, and recommend what to do.
• They lack visibility. Most contacts go unreported. Trucks come back with a small ding, and nobody knows where it happened. Managers walk the aisles and find damage on shelves and product.
• Pedestrians are only part of the risk. Hard braking, hard cornering, collisions with racking, and OSHA violations matter as much as a person in a blind spot.
• Noise kills trust. A forklift moving forward while people walk away should not trigger an alarm. Every false positive costs attention the manager does not have.
• It has to fit their systems. Findings must flow into the tools and reports they already use, so they can go up the chain and change behavior.

What I did

• Ran discovery. Found and interviewed EHS managers, mapped their current process, pain points, and what visibility they lacked.
• Shaped the detection scope with our industrial engineer. From pedestrians only to unsafe driving, collisions, and violations, using IMU signals and video.
• Simplified installation. Worked from mounting schematics to a self-install kit with recommended positions, so a customer can go live without us on site.
• Pitched, converted, and ran pilots. Met customers regularly, walked them through their analytics, and turned their feedback into the next iteration.
• Tuned signal against noise. Severity ranking and motion awareness, so managers see about one serious event a day instead of a feed of alarms.
• Shipped the safety analytics dashboard to production with Claude Code. Data visualization, event review, and the workflows managers use to confirm incidents, tag them, and feed the model.

How the product changed

• Pedestrians in blind spots→Driving behavior, collisions, and violations
• An alarm on every detection→Severity-ranked events, about one serious a day
• Raw footage→Curated events with recommendations
• Installed by us, on site→Self-install kit, under two hours
• Real-time alerts for the driver only→Every incident logged, so safety managers can review it and find systemic risk

Gallery

Canary mounted on a forklift in a warehouse

The same forklift with the two cameras, the nest, and the display outlined

How Canary mounts on a forklift: camera 1, nest, camera 2, and the driver display

Recommended camera positions

Recommended nest position

Recommended display position

Early prototype: a webcam on a mount

Early prototype: the alarm light

Early prototype: the all-in-one unit with display

Camera mount on the overhead guard

The nest mounted on the post

Driver display in the cab

Early dashboard prototype: near misses by time of day

Early dashboard prototype: key safety indicators

Incidents per hour, minutes of operation, and near-miss risk score

Near-miss score and weekly trend

Hard braking events with severity and pedestrians detected

Event review: confirm, tag, and resolve

Safety analytics dashboard

A pilot site: forklift with Canary in a warehouse

Where the company was

• The platform had scaled. 30,000+ monthly active users, a white-label factory, and 135+ branded apps on the way.
• The constraint moved from building to delivering. It took about four months to take a customer live and about three months before they saw a beta.
• Requirements collection cost customers 6 to 12 hours of their own time, and churn sat around 25 percent.

The problem

• Customers waited months for value while implementation, design, and app-store setup ran through engineering.
• Every implementation was custom. Inventory files, branding, app-store accounts, domains, and design approvals arrived in a different order for every customer.
• Nobody had one view of the customer. Usage, requests, support tickets, and health lived in different tools.
• Churn was found late. Risk showed up at renewal, not when it started.

Who it is for

Everyone between a signed contract and a renewed one:

• Customers implementing their apps
• Customer success managers
• Implementation and support
• Sales
• Product

What I did

• Designed the five-phase onboarding pipeline. Sales, Discovery, Implementation, Beta Testing, Launch. Each phase with a defined outcome and standardized requirements collection, so customers knew what to bring and when.
• Built the internal setup and admin tools so customer success could configure organizations, inventory, apps, and custom designs without waiting on engineering.
• Automated the recurring implementation work and moved design approval into comments instead of meetings.
• Built the customer portal and help bot. Customers submit and track requests, get answers, and configure their own apps.
• Integrated the support infrastructure into one view of the customer. Rocketlane for delivery, Salesforce Service for omnichannel support and a knowledge base, health scores and early-warning alerts on top.
• Ran quarterly business reviews and turned onboarding friction and health signals into product priorities.
• Applied the same system to the pre-sale journey. Rep-by-rep prospecting became a structured Salesforce funnel with a dedicated SDR role and automated outreach. Pipeline that had been flat quarter over quarter grew about 20 percent a quarter, at every stage from cold call to deal.

Impact

Gallery

Slides from the Customer Success town hall, 2022 Q3, and the analytics I presented to the team. Customer data removed.

The onboarding pipeline: Sales, Discovery, Implementation, Beta Testing, Launch, each with its outcome

The starting point: about four months to go live, 6 to 12 hours of customer time to collect requirements

What Discovery collects before implementation starts

The customer success pipeline as a flow

Accounts in each onboarding phase, 2022 Q3

Presenting the executive summary of plan usage to the team

The early-warning view: usage signals per account over the last 30 days, account names left out

How to reduce churn: one view of the customer, early warning, business reviews

Customer health score inputs

Customer health score bands and the action for each