Work

Twenty years of shipping
the real thing.

Global-brand work that proves the range (each brand its own case, not a logo soup) and own builds that prove it's production-grade right now. The own builds are 100% mine, so each one spans all four verticals: product, tech, strategy, design.

Built by hand, production-grade, now

Own products · 100% solo · all four verticals each
Placeholder Architecture / knowledge-graph view
Frontier build · architectural complexity

SymbiOS

A multi-agent system for systemic therapy: a per-user Guardian AI, a consent-gated knowledge graph, anonymisation before any model ever sees data, and secure transfer of highly sensitive information. Around 70% of replies served locally without an LLM call to keep cost and latency down, with a clear path to faster inference and a semantic cache as it grows. Roughly 56k lines of TypeScript.

TechProductStrategyDesign
Placeholder Dashboard / widget screenshot
AI sales platform · in production

autoply

Multi-tenant AI sales SaaS: tenant isolation, plan-based quotas, 18 parallel workers, model routing per task (frontier where output value is high, lean for everyday queries), an embeddable widget and a curated knowledge layer across 20 verticals. In production today on AWS cloud, run cheaply at small scale, with a deliberate distributed-per-market architecture ready to scale. Roughly 2 months.

TechProductStrategyDesign
Placeholder Gallery / similar-images panel
Photo SaaS · collaborative curation

Vuelette

A complete photo curation and delivery SaaS (auth, galleries, ratings, sets, public share links) with in-process visual vector search (cosine in Node, no vector DB) that finds similar images cross-shoot, and an in-browser healing tool. Runs on SQLite today at small scale, with a clear roadmap to Postgres and beyond. Built end to end.

TechProductStrategyDesign
Placeholder Terminal / graph / curation UI
AI power-user tools · used daily

AI Power-User Tools

We're in a new paradigm, and the tooling to work at very high level and speed with AI doesn't exist yet, so I built it. hmux (a native macOS terminal for running tens of Claude Code sessions in parallel), AutOS (a code-intelligence metadata layer over MCP, 66 to 76% fewer tokens per query), and Nodio (a voice-to-knowledge-graph synthesis of both). The leverage that makes one-person rebuilds possible.

TechProductStrategyDesign

Every own build carries one story: it runs extremely cheaply today at small scale, and there's a clear plan for how it scales, keeping costs down so you never pay for more than you need when you need it. The builds run by request only, early stage on purpose: a select set of clients, all the testing done before we scale. The opposite of throw-it-out-and-crash.


Across the world's biggest brands

Each brand, its own case · the disciplines I owned
Google
Dentsu · Google Sweden account lead

Google

// The problem

A flagship global account that needed to ship live digital products and broadcast technology in-market, not creative concepts that died in a deck.

// What I did

Led the Google Sweden account and grew it 350%. Ran Google Hangouts Live broadcast technology hands-on, the only team delivering it in Sweden at the time, alongside the strategy and product work behind the account.

// Result

A tripled account and live broadcast tech in production. Strategy, tech and product carried together by one lead.

StrategyTechProduct
Mondelēz
Dentsu · enterprise platform

Mondelēz

// The problem

A global FMCG brand needed a loyalty platform that actually ran, with recurring engagement, real users and real data, not a campaign microsite.

// What I did

Built the Gevalia loyalty platform end to end: the product thinking, the architecture and the experience. A production system holding real members, not a one-off activation. It launched and it worked.

// Result

A live loyalty platform for one of the world's largest brands, owned across product and tech, built to hold real members, not run as a campaign.

ProductTechStrategy
Procter & Gamble
Dentsu · product concept

Procter & Gamble

// The problem

P&G wanted to reach households inside daily routine, not interrupt it with advertising. A product people would actually use, with the brand woven in naturally.

// What I did

Shaped a chores and household-management app concept: the product proposition, how families would use it day to day, and where the brand lived inside the experience rather than on top of it. Concept and product design, made to be built.

// Result

A product-led brand concept for a global FMCG leader. Proof the range runs from packaging to platforms to product ideas.

ProductDesignStrategy
Panasonic
Dentsu · real-time platforms

Panasonic

// The problem

Olympics-scale moments demanded real-time digital platforms that stay up, stay fast, and update live under heavy, spiky traffic.

// What I did

Delivered Panasonic real-time Olympics platforms: products built for live data and burst load, where the architecture has to hold at the exact moment everyone shows up at once.

// Result

Real-time platforms shipped for a global brand at event scale, the kind of load most demos never face.

TechProductStrategy
Candy People
Brand + product build · Global VP

Candy People

// The problem

A confectionery business that needed a portfolio with real shelf presence: distinct brands, distinct products, distinct packaging, not one me-too line competing on price.

// What I did

Thirteen brands and 80+ products, end to end: brand positioning, packaging design, and the go-to-market behind each one. Product strategy, design craft and commercial execution in a single mandate.

// Result

A differentiated portfolio that scaled the company hard. The revenue (167%) was the consequence, not the headline.

ProductDesignStrategy
Kverneriet
CTO · CIO · CMO

Kverneriet

// The problem

A multi-brand operator running on disconnected tools, with no shared platform and no operational intelligence. Then COVID hit and the model had to change overnight.

// What I did

Built, across three executive hats (CTO, CIO, CMO): a multi-tenant operations platform plus an AI operational-intelligence layer. A cost-calculation system took pricing iterations from roughly 2 weeks of full-time work down to about 45 minutes, enough to reprice monthly instead of 2 to 3 times a year (a potential ~24 weeks of work saved annually). An inventory system turned a full-day job for 5 to 6 people into a couple of spread-out hours (~360h saved per year). And around 7 to 8 months into COVID I took over and re-did the entire digital flow plus the marketing.

// Result

+500% revenue on home delivery and takeaway within weeks of the COVID pivot. The systems shipped and proved enormous time savings: weeks of pricing work down to minutes, and a full day of inventory work down to a couple of hours.

TechProductStrategy
Strategic Design
Own studio · two stints, 2003 to 2011

Strategic Design

// The problem

Blue-chip clients needed work that connected visual and system craft to business strategy: brand and product that differentiates, plus the technical and process backbone to deliver it.

// What I did

Ran my own studio across two stints over many years. Beyond branding, UX and packaging: system design, system concepts and system architecture, development project management, and digital due diligence, for clients including Absolut and Ubisoft.

// Result

A decade-plus of brand, product and system work for demanding clients. The design and architecture root everything since is built on.

DesignStrategyTechProduct

Have a product in this shape?

If any of these stories sound like your codebase (proved the market, now fighting its own architecture) that's exactly the work. Let's compare notes in 20 minutes.