The work I’ve done, and what I’m building on my own.
The career work at Smart Food Safe and Toluna, in depth. Plus two learning projects I use to sharpen the same skills on my own time.
The work, in depth
Two companies, the work behind the highlight reel: what the role needed, how I approached it, what changed.
Brand Marketing Manager
Building an end-to-end lead-attribution system
Marketing needed one reliable view of which channels and content were actually driving pipeline.
The function was scaling quickly across several international markets, and lead sources needed to be tracked in one place rather than across separate tools.
Design and implement a centralised CRM and attribution setup end to end, rather than layering another tool on top.
Implemented the CRM, mapped every lead source, set up clean, auditable reporting the team could actually trust.
Clean, auditable lead-source reporting, giving the team a solid basis to double down on what works.
Repositioning the brand for an enterprise audience
The product had grown well beyond how it was being described, and the messaging needed to catch up with what enterprise buyers care about.
The visual identity and voice had not been formally documented, so different teams were interpreting the brand slightly differently.
Reposition end to end: a new visual identity, a written voice-and-tone playbook, and messaging rebuilt around buyer outcomes like audit readiness and operational visibility, not product features.
Led the repositioning across identity, website and product messaging, working with design and content to apply it consistently across every customer touchpoint.
A consistent, enterprise-grade brand that reads credibly to large multinational buyers.
Building an AI-assisted content engine the team could actually run
A content engine spanning several international markets needed to move faster than a small team could manually sustain, without dropping quality.
Content, design, SEO and video work were each running as separate manual processes without a shared playbook.
Introduce an AI-assisted content workflow paired with a documented operating playbook, so AI speeds up production without lowering the quality bar.
Rolled out the workflow and playbook across the content team, then extended the same systems thinking to multilingual go-to-market and international trade-show marketing.
A content engine that moves faster without losing enterprise quality standards, run by a small team across several international markets.
Jumpstart Research Manager, Consumer Insights & Brand Strategy
Turning global research into decisions Fortune 500 teams actually acted on
Clients were getting research data, not decisions. Insights sat in decks without a clear path to action.
Global consumer-intelligence work across usage and attitude, brand tracking, advertising effectiveness and product innovation, in 70+ markets.
Build a cross-study synthesis framework that surfaces white-space opportunities instead of reporting study by study.
Applied the framework at scale, including a large product-testing programme for United Breweries.
Informed real product-portfolio decisions for FMCG clients, cut Power BI reporting turnaround by 25%.
Standardising client reporting so it did not depend on one person
Client reporting was manual and slow, and the growing analyst team had no shared workflow for producing it.
20+ concurrent research programmes each needed recurring reporting, rebuilt from scratch by whoever picked up the job.
Standardise reporting into reusable Power BI dashboards, then teach the workflow instead of keeping it as one person’s tribal knowledge.
Built the dashboard templates, automated the manual reporting steps, and mentored three junior analysts on the analytics workflow and the brand-positioning frameworks behind it.
Client turnaround time on reporting fell by 25%, and the team could run the workflow without bottlenecking on one person.
What I build to learn
An AI marketing systems platform for small B2B teams
Most small marketing teams adopted AI tool by tool. Output went up. Strategy did not, and nothing underneath it was repeatable.
Free, plain-English frameworks plus a content gate: a written quality standard every piece has to clear before it ships, so speed never costs credibility.
I wrote the brand rules, the voice, the quality bar and the publishing cadence once, as instructions a machine can follow. An AI system then researches, drafts, checks its own work against those rules, publishes to the site and schedules the social posts. I review direction and standards, not individual tasks.
With me. The AI is fast and consistent and has no taste. Positioning, what is worth saying, what gets rejected and when the rules themselves are wrong are all still mine. The interesting work moved up a level rather than disappearing.
Founders, marketing managers and small B2B teams who want a system instead of another tool to manage.
That one person with a clear system can run a real publishing operation end to end, and that the constraint on AI marketing is the quality of the instructions, not the model.
A verdict-first review site for home-office gear
Most gear-review sites either imply hands-on testing they never did, or bury a straight answer under SEO padding.
Verdict-first reviews built from published lab testing, owner reports and spec sheets, with one hard rule written into the system: never imply testing we did not do.
Same architecture as the other project, different rulebook. I set the category strategy, the honesty rule and the scoring method; the AI system does the research, drafts the verdict, produces the short-form video and schedules it. Deliberately built as a second brand so the underlying system had to prove it transfers.
Choosing which categories to enter, where the credibility line is, and when a sourced claim is still not good enough to publish. The honesty constraint is a strategic asset here, not a limitation, and that call is a human one.
Remote workers and home-office buyers deciding on monitor arms, desks, chairs and the rest of a setup.
That the same operating system runs a B2C review brand and a B2B education brand without being rebuilt, which is the difference between a workflow and an actual system.
How I Run These With AI
Promptly Market and GizmoJudge don’t run on hours I don’t have. I built each one on a set of rules once, and an AI system executes them day to day: drafting posts, checking every piece against that site’s own brand and quality standards before anything goes out, publishing on a schedule, and posting to each site’s social channels without me opening a dashboard.
I set the direction and the guardrails. The system does the repetitive part: writing, checking its own writing against the rules I gave it, publishing, and posting to social. I check in on the results, not the busywork.
It’s not perfect. I’ve caught and fixed real mistakes along the way: a missing image, a stalled routine, one page saying something another page didn’t. Each one made the rules a little sharper. Build a system, watch where it breaks, fix the actual cause instead of the symptom. That’s the same loop I’d bring to any brand I ran.