Building Refinitiv Eikon's cloud content platform for a thousand institutional desks across thirty-five markets.

Refinitiv (then Thomson Reuters) had a content estate sprawling across 260+ tools and ingestion paths. Different feeds, different schemas, different teams, the same downstream user trying to find a single number on their terminal. Cloud migration was the obvious enabler — but the harder question was content operations. Manual content specialists were the bottleneck on every new attribute, every new market, every new language pair.

The product opportunity wasn't "move to AWS". It was "redesign the operating model so the platform scales when the markets do". That meant pairing the cloud platform build with an ML/NLP content automation programme — and proving institutional users got the same data quality they were paying for.

• Ran discovery across 260+ existing tools and ingestion paths to define the platform strategy, target architecture, and prioritised backlog. Built the case for what to migrate, what to consolidate, and what to retire — grounded in usage data and divisional dependency mapping, not architectural preference.
• Defined the cloud platform product on AWS — content ingestion, normalisation, search, delivery to Eikon front-end, and the analytics layer for institutional users.
• Designed an ML/NLP content extraction pipeline pulling 250+ structured and unstructured attributes from raw data feeds. Operational across 60+ language pairs. Confidence-scored outputs gated by human review for the highest-stakes attributes.
• Re-engineered the content operations model. Scaled from 208 manual content specialists doing direct extraction, to 10 specialists overseeing automated processing at 90%+ accuracy — with the team's role redefined from production to quality assurance and edge-case adjudication.
• Owned the institutional user backlog. Prioritised features against a base of 1,000+ active institutional users across 35+ global markets, balancing front-office demand for new attributes against platform investment.
• Defined and tracked platform KPIs — content coverage, freshness, accuracy, search relevance, user engagement — reporting to senior Refinitiv leadership and content-domain stakeholders.

The pattern that made this programme work — and the one I've carried into every senior role since — was treating the content operations team as a stakeholder, not a cost line. The 208-to-10 scale shift sounds dramatic, but the team didn't disappear; they were redeployed into the quality assurance and edge-case roles that the automation needed. The senior content specialists became the auditors, trainers, and adjudicators of the ML pipeline they used to feed.

That framing — automation as augmentation, headcount as expertise — is what kept the programme moving. It's the same instinct I carried into GenAI deployment at Deutsche Bank two years later.

Demonstrable AI platform transformation with measurable operational impact — and a content operations model that scaled with markets, not against them. The technical patterns and operating model from this programme inform how I now think about AI deployment in regulated finance.