Professional

Choreograph is a spinout startup formed from joint WPP and Google initiatives, focused on enterprise AI, customer data and media investment optimization. The work sits in a financially important part of the media business: helping major brands understand audiences, allocate advertising budgets more intelligently and improve commercial outcomes from large-scale data.

The role carried broader AI/ML responsibility across data science, machine learning and engineering workstreams while staying close to model direction, architecture and implementation quality. The work involved multimodal AI, synthetic data, representation learning and large-scale customer understanding from surveys, text, video, web-scale signals and behavioral data.

This role began when Choreograph was around 50 people, before it scaled toward roughly 1,000. Compared with earlier startup roles, the environment was larger and more complex: enterprise customers, larger data systems, more coordination and higher pressure to make models robust, explainable and commercially useful.

Yupana was an early-stage financial SaaS startup focused on automating financial-management workflows. The role established the company's first data-science function at a point when much of the automation logic was still rule-based, moving the product toward machine-learning-driven document intelligence and workflow automation.

The work covered first ML architectures, early model versions, data pipelines, service design and MLOps foundations. It also included interviewing and hiring technical people, defining how the data-science function should operate and connecting model development with production software.

The company later became part of Cegid, one of Europe's major financial SaaS and business software companies with roughly 3,000-plus people. The role was a clear zero-to-one startup setting: roughly 20 people, first ML capabilities, real customers, fast commercial growth and an eventual path into a larger European software platform.

Retail Insight was a financial-optimization SaaS company for large retailers and CPG partners, using machine learning and decision science to turn store-level operational data into margin, availability and waste-reduction decisions.

The role involved early responsibility for developing new product initiatives from concept into working software with direct contact with large retail customers and ownership over technical delivery. The work translated operational and financial retail problems into forecasting, anomaly detection, optimization, data-quality and decision-support systems.

The company later scaled from roughly 30 people into a much larger retail-optimization software business of around 250 people, making this an early exposure to high-growth startup execution: small teams, large customers, commercial pressure and ML systems that had to improve real operating decisions.