Selected Work

Products that moved the needle.

I own the supply side of a marketplace: the APIs, listing systems, and data pipelines behind what buyers search, find, and trust. Four problems I took on, and what changed.

Flagship Case Study

Re-Architecting a Partner API Ecosystem Serving 7,000+ Dealers

A fragile integration layer carrying 250,000+ listings a year, rebuilt without breaking buyer-facing data quality or dealer trust in the transition.

40%

Reduced manual entry time

55%

Increased API adoption

20%

Higher structured data completeness

10+

New buyer search filters enabled

30%

Faster cross-functional decisions

250,000+

Listings via partner APIs (2025)

7,000+

Unique dealers (2025)

99.99%

Partner API uptime

The Challenge

Our marketplace's dealer ecosystem depended on fragmented legacy integrations. The API layer was mission-critical, powering thousands of listings and feeding the structured data that buyer-facing systems (search, recommendations, listing pages) depend on.

The system was brittle and hard to scale, with every new partner adding onboarding friction and support load. A feed format mismatch or missed endpoint change would surface not as an alert, but as incomplete listing data visible to buyers, sometimes for hours before anyone noticed. New partner integrations required manual intervention and could take weeks to go live.

What I Did

Designed the Next-Gen API Platform

Led a full re-platforming of partner-facing APIs, shifting from manual entry dependency to structured data-driven ingestion for 7,000+ dealers across the ecosystem. We chose to build a new structured ingestion layer rather than patch the existing XML feeds. A longer initial runway, but the only path to data completeness that buyer-facing search and recommendations could reliably depend on.
Owned Reliability as a Product Metric

Treated uptime, data consistency, and integration tickets as first-class product KPIs, with observability and adoption tracking across endpoints.
Orchestrated Enterprise Migration

Aligned engineering, product support, sales, and key accounts to reduce rollout risk, managing system transition without disrupting marketplace liquidity or the buyer-facing data quality downstream teams depend on. We ran both systems in parallel during migration rather than cutting over hard, accepting months of operational overhead to protect marketplace liquidity.

Impact

Converted a fragile integration layer into a scalable partner platform. Structured data completeness improved by 20%, enabling 10+ new buyer search filters with over 80% data coverage across the marketplace.

API StrategyPlatform ArchitectureB2BEnterprise Migration

Case Study 02

Scaling the Listing Platform to Increase Liquidity and Dealer Adoption

In a marketplace, speed equals liquidity. Cutting listing creation time from 15 minutes to 6.5, and building the experimentation rails to measure downstream effects.

35%

Faster time-to-market

Impact

Increased inventory velocity and seller efficiency. Higher listing completeness reduced listing page bounce rates by 34% and improved search relevance for buyers across the marketplace.

Marketplace LiquidityExperimentationAutomationData Quality

Read the full story Hide the full story

20%

Dealer adoption increase

18%

Listing completion rate

12%

Reduced seller drop-offs

30%

VIN lookup adoption

45%

Fewer manual data errors

The Challenge

Dealer workflows were slowing inventory velocity. Listing creation was manual, slow, and inconsistent.

With no VIN lookup, a dealer listing a vehicle entered every attribute by hand: make, model, trim, engine spec, field by field, averaging 15 minutes per listing. Structured data was incomplete enough that key buyer search filters could not be reliably enabled at scale.

What I Did

Automated Core Listing Workflows

Scaled structured listing infrastructure, cutting average listing creation time from 15 minutes to 6.5 minutes and improving ingestion quality.
Built an Experimentation Engine

Designed a structured A/B testing framework embedded into listing workflows. We chose to build a lightweight framework directly into the listing flow rather than wait for a platform-wide testing tool, trading some statistical sophistication for faster iteration.
Strengthened Vehicle Data Services

Optimised VIN identification and vehicle data services used by external partners, increasing structured data adoption and reducing manual errors.

Case Study 03

Launching AI-Powered Listing Descriptions in a Marketplace Core Flow

Building a generative AI product from scratch inside the marketplace listing flow, with no prior AI infrastructure and no validated demand.

22%

Increased listing engagement

Impact

Proved that better seller-side content lifts buyer engagement and lead conversion. One product moved both sides of the marketplace.

0→1 ProductGenerative AIA/B TestingConversion OptimizationContent Automation

Read the full story Hide the full story

15%

Higher lead conversions

<2 min

Description time vs. ~12 min manual avg

The Challenge

Dealer-generated listing descriptions were inconsistent, low quality, and performance-sensitive. Descriptions written manually ranged from a single sentence to copy-pasted marketing text, with no consistency and limited correlation to actual buyer engagement. There was no AI infrastructure in place; we lacked generative workflows, a clear adoption path, and an experimentation baseline.

This was a zero-to-one product with no baseline to lean on. It shipped without validated demand, and without any guarantee dealers would trust automated copy or that engagement would hold.

What I Did

Framed the Bet Up Front

Defined the product bet upfront: if we automate structured listing descriptions aligned with vehicle attributes, we should increase engagement and lead conversion while reducing seller effort. We scoped v1 to AI-suggested descriptions only, with no dealer editing or customisation controls. Stakeholders wanted more dealer control from day one; we pushed back to isolate the quality signal before adding complexity.
Built the Experimentation Layer First

Before scaling AI output, designed success metrics, defined quality guardrails, structured A/B validation, and ensured reversibility. We delayed go-live to instrument measurement properly. Shipping first and measuring retroactively was the path of least resistance, and we did not take it.
Embedded Into Workflow With Minimal Friction

Integrated AI suggestions into listing creation rather than forcing review-heavy editing flows.

Case Study 04

Building Mission-Critical HR Data Infrastructure for a Global Enterprise

Designing the HR data backbone for a global pharmaceutical enterprise, where a single data error can mean a missed paycheck or a compliance incident.

40%

Process efficiency improvement

Impact

Transformed fragmented HR systems into a unified data platform that powers payroll, compliance, and workforce operations across the enterprise.

Enterprise PlatformAPI InfrastructureData GovernanceRegulated Environment

Read the full story Hide the full story

Business units on one platform

Compliance incidents post-launch

30%

Faster downstream development

20%

Faster release cycles

40%

Fewer post-launch issues

The Challenge

A global pharmaceutical enterprise relied on fragmented, legacy HR systems. Payroll, compliance reporting, and cross-department data access were bottlenecked by inconsistent data flows and manual processes. A wrong cost centre or missed employee record could delay a paycheck or trigger a compliance audit in markets with strict employment law.

The organisation needed a unified data backbone: core HR APIs and automated pipelines that give business units real-time access to governed, audit-ready data.

What I Did

Built Core HR APIs & Data Pipelines

Designed and delivered a suite of enterprise-grade APIs that standardised how HR data was accessed across payroll, workforce planning, and compliance, replacing fragmented manual processes with reliable, governed data services.
Automated Data Access & Governance

Implemented automated data provisioning and governance controls, reducing manual data requests and ensuring consistent, auditable access across business functions in a regulated environment.
Defined the Product Roadmap Across Competing Workstreams

Owned prioritisation and trade-off decisions across a CHF 2.5M investment spanning multiple workstreams. We sequenced payroll and compliance data first, delaying workforce analytics features that had stronger stakeholder demand. In a regulated environment, the wrong sequence means audit exposure, not just delayed features.

Have a platform problem like one of these?

If you're untangling a fragile integration layer, scoping a 0-to-1 bet with no baseline, or building a data platform people have to trust, that is the work I do best. Tell me about yours.

Let's chat LinkedIn