Hi, I'm Robin

A strategist and AI builder who'd rather build the idea than sit through the meeting about it.

About me

For seven years, I've worked on AI platforms and growth strategies for major brands, including L'Oréal and Forvia. I turn business problems into AI systems designed to scale. I trained in strategy, and I like building things myself.

Work

Self-built AI product

Dionysus

Born from a passion for cocktail craft and building things from scratch. An immersive, sensory web experience where three AI agents craft bespoke cocktails and stories for 132 distinct personas.

  • Agentic Development
  • Multi-Agent Studio
  • Interactive Cinema

132 bespoke personas & cocktails

Launch live app

Self-built AI product

2026 Trends Atlas

A RAG-powered knowledge graph built from 70+ trend reports. Ask questions about 2026 trends and trace the answers back to their sources.

  • LLM-Wiki Architecture
  • Knowledge Graphs
  • Trends 2026

70+ trend reports ingested

Launch live app

Self-built AI product

Sigma AI

An autonomous platform where AI agents turn research into strategic insights and presentation-ready decks in minutes.

  • Multi-Agent Systems
  • Prompt Engineering
  • n8n

18 MECE sub-briefs per run

Launch live app

AI platform · Product design

Power Claims

An AI platform for L'Oréal's claims process, from creating and substantiating product claims to using them in local markets.

  • AI Platform
  • Product Design
  • Process Reinvention

130+ champions trained on the new process

GenAI · Innovation

Forvia-X

Mapped 60+ GenAI use cases across Forvia's innovation process and built proofs of concept for the 10 with the highest value.

  • GenAI
  • Process Design
  • Value Case
  • Roadmap

60+ use cases mapped

AI project

SongAI GenAI platform

Led product strategy for Accenture Song's global GenAI platform, bringing research, strategy, content, and delivery tools into one modular system.

  • GenAI Strategy
  • Marketing reinvention
  • Platform design

70+ practitioner interviews

Metaverse · Gaming

Quest for Hope

A decentralized conservation campaign on Roblox and Decentraland, where players meet digital twins of endangered Arabian leopards. Launched on the first UN International Arabian Leopard Day.

  • Roblox
  • Conservation
  • Gaming
  • NFT

19.1M press views

Launch live app

Let's build

Contact me

Self-built AI product

Dionysus

Self-built · Joining creator craft with the history of cocktails

Summary

Dionysus brings together two of my biggest passions: building and bringing things to life as a creator, and my longstanding obsession with the craft and lore of cocktails.

For years, I've loved creating personalized cocktails for people, drawing inspiration from the history of rare ingredients, legendary bartenders, and who made what for whom. But bringing that personal magic to the web faced a classic problem: recommendation questionnaires are boring. Nobody wants to fill out clinical forms or click through dry surveys. I wanted an experience that was tactile, cinematic, and genuinely fun to explore. Using modern agentic development and BMAD methodologies, I built the entire experience end to end on my own. In the front, visitors embark on a five-minute dark-room descent set to sumi-e ink washes and generative audio, making instinctual choices through physics interactions like torchlight reveals, drifting gravity motes, and cooling embers. Behind the curtain, three specialized AI agents (The Psychologist, The Historian, and The Mixologist) collaborate in an authored Pour Studio to craft 132 bespoke cocktails and personal stories tailored to every unique persona.

Key outcomes

132
personas with bespoke cocktails & stories
3
Pour Studio specialist AI agents
1
creator building end to end
0
boring form fields or checkboxes

What I did

  • Conceived, designed, and engineered the complete product solo from scratch, applying product strategy, interactive design, and agentic development workflows
  • Reinvented the recommendation questionnaire: replaced sterile form inputs with a sensory, five-minute dark-room journey featuring custom canvas physics (torchlight reveals, liqueur seeds, drifting gravity motes, cooling embers) synchronized to a continuous video journey and generative soundscape
  • Architected the multi-agent Pour Studio: designed three collaborating specialist AI agents: The Psychologist mapping personal temperaments across 132 archetype pairings, The Historian researching cocktail provenance and bar lore, and The Mixologist formulating balanced drink specs
  • Encoded years of personal cocktail craft, spirit histories, and classic bartending philosophy into structured agent workflows so every cocktail has genuine character and provenance
  • Built a deterministic matching and guardrail engine that routes instinctual user gestures to one of 132 authored persona dossiers with zero hallucinations
  • Leveraged BMAD methods and agentic AI tooling to deliver full-stack code, UX design, custom interactive components, and production deployment on Vercel as an agile solo creator
  • Directed and produced the cinematic teaser trailer with HyperFrames, pairing headless browser captures of the interactive rituals with custom sound design

Business impact

  • Proves how agentic development and BMAD methods empower a solo builder to design, architect, and ship a world-class, production-grade product from concept to launch
  • Solves the survey-fatigue problem by turning a routine recommendation quiz into a high-retention, cinematic ritual people genuinely enjoy completing
  • Demonstrates how Generative AI can scale nuanced human craft, storytelling, and domain expertise without flattening creativity or losing authenticity
  • Provides an inspiring blueprint for brands looking to replace generic chatbots and forms with memorable, emotionally resonant product discovery experiences

Competencies

  • Agentic Development
  • Multi-Agent Systems
  • Product Design
  • Creative Direction
  • Interactive Cinema
  • BMAD Methods

Visit live product

Self-built AI product

Self-built · Trend research and knowledge graph

Summary

The Atlas brings together 70+ trend reports from Accenture, Gartner, Deloitte, McKinsey, WGSN, GWI, and other publishers. Each report lives in a separate PDF, so comparing their findings takes manual reading. I built the Atlas to keep the connections between reports available as new sources are added.

I used Andrej Karpathy's "LLM-Wiki" pattern to build a persistent, interlinked wiki between the user and the raw reports. As each source is added, an LLM extracts its claims, files them into structured pages, and updates the cross-references and synthesis. The wiki keeps that knowledge available for future questions and grows with each source. I built an interactive graph and a query layer on top: ask a question in plain language, and the Atlas reads its index, follows links to relevant pages, and returns an answer with claims mapped to source evidence. The system works without a vector database.

Key outcomes

70+
trend reports ingested
17
cross-domain concepts mapped
0
vector DBs (index-based retrieval)
1
compounding knowledge graph

What I did

  • Designed and built the complete Next.js application and deployed it on Vercel
  • Implemented the Karpathy LLM-Wiki pattern: a three-layer architecture of immutable raw sources, an LLM-maintained markdown wiki, and a schema file governing ingest and synthesis conventions
  • Built the ingest workflow so the LLM extracts claims from each report, files them into structured source/concept/entity pages, and maintains cross-references and contradiction flags across the wiki
  • Built index-based retrieval: the query engine reads the content catalog (index.md), then follows links to relevant pages. With 70+ sources, I chose to work without embeddings or a vector database
  • Built the graph visualization layer from the wiki's Obsidian-style links: nodes, weighted connections, type filtering, search and navigation
  • Built a query engine with two modes: deterministic retrieval and ranking, and an optional LLM-grounded mode via OpenRouter
  • Added inline citations that map each answer's claims to source pages so readers can check the evidence

Business impact

  • Keeps a reusable knowledge layer that grows with each source added
  • Finds connections across reports and research domains
  • Turns days of manual report reading into minutes of querying for strategy and innovation research
  • Links synthesized answers to their evidence for use in client work and presentations
  • Provides a repeatable way to turn a collection of sources into a searchable graph and an LLM-maintained wiki

Competencies

  • LLM-Wiki Architecture
  • Knowledge Graphs
  • Trends 2026

Visit live product

Self-built AI product

Sigma AI

Self-built · Research and insight generation with AI agents

Summary

Turning research into insights that change a decision takes time and depends on a consultant's judgment. I built Sigma AI to automate more of that work.

The autonomous platform takes a company and a research brief, splits the topic into 18 MECE sub-briefs, and assigns them to AI research agents running in parallel. Findings are organized using the proprietary SCHEMA framework's six lenses: Structures, Culture, Humans, Economics, Machinery, and Actors. Comparing pairs of lenses, such as Culture × Economics, helps identify tensions and opportunities across the research. Each insight is scored against FPET (Fresh, Potent, Energizing, True). An automated review loop critiques and rewrites weak insights until they meet the quality threshold. Sigma AI automates a consulting research workflow and produces presentation-ready insight decks.

Key outcomes

18
MECE sub-briefs per run
6
SCHEMA research lenses
4-point
FPET quality scoring
End-to-End
research → deck pipeline

What I did

  • Designed the architecture for AI agents to coordinate research and produce insight decks
  • Developed SCHEMA, a proprietary framework that organizes research across six strategic dimensions
  • Engineered the orchestration logic: MECE decomposition into 18 sub-briefs, parallelized AI research agents, and SCHEMA-based consolidation
  • Built an insight engine that compares pairs of dimensions to identify tensions and opportunities
  • Created the FPET evaluation system (Fresh, Potent, Energizing, True) and the autonomous QA loop that scores, critiques, rewrites and version-tracks insights until quality thresholds are met
  • Built n8n pipelines using OpenRouter, GPT models, and Perplexity to run research and synthesis in parallel
  • Automated the final deliverable: Google Sheets and Google Slides integration for templated, presentation-ready insight decks

Business impact

  • Automates research-to-insight synthesis
  • Applies a consulting methodology through a repeatable workflow
  • Runs research tasks in parallel, combining days of analyst work into a single run
  • Scores, critiques, and revises insights automatically through the FPET review loop
  • Produces presentation-ready decks through automated research, synthesis, and slide generation
  • Combines strategy, AI systems design, and foresight research in one workflow

Competencies

  • Multi-Agent Systems
  • Prompt Engineering
  • n8n

Visit live product

AI platform · Product design

Power Claims

L'Oréal · Global Claims Transformation

Summary

Claims on L'Oréal's packs, web pages, and ads passed through separate tools and manual handoffs between teams. Indie brands and K-Beauty competitors were releasing bolder claims faster.

I led product design and AI strategy for Power Claims. The platform connects competitive intelligence, brief creation, scientific substantiation, regulatory review, and local activation. The skincare MVP launched across eight pilot countries. AI agents coordinate work across five modules, with workflows for global marketers, regulatory reviewers, and local teams.

Key outcomes

130+
champions trained on the new process
11
countries in Claim Committee rollout
8
pilot countries at MVP launch
3
skincare pilot projects live on the platform

What I did

  • Sole design and AI strategy lead, from vision through MVP delivery and the post-MVP roadmap
  • Designed the end-to-end claims lifecycle architecture across five modules: Claim Knowledge, Claim Project, Predict Claim, Risk Escalation, and Claim Dossier
  • Designed 20+ AI agents, from an orchestrator that drafts full claim lists to specialists scoring consumer relevance, competitive gaps, scientific grounding, and regulatory risk
  • Prototyped the complete claims pipeline with AI agents before platform engineering began, testing the architecture with real dossiers and users
  • Designed workflows for five user roles, including marketers, regulators, and scientists, so they could work together in one tool and focus on tasks relevant to their roles
  • Connected competitive and consumer intelligence (Mintel, Revuze, TalkWalker, internal insight platforms) so teams could use market evidence when developing claims
  • Defined the scaling roadmap: integration of AI capabilities, broader process coverage, and expansion across categories and divisions

Business impact

  • Selected as L'Oréal's global platform for claims transformation, replacing separate legacy tools
  • AI agent architecture adopted as the reference model for L'Oréal Beauty Tech's AI platform
  • 130+ R&I and marketing champions trained, 3 pilot projects onboarded, rollout underway in 11 priority countries
  • Defined expansion across L'Oréal's divisions through 2027, with goals of bolder claims, shorter time to claim, and better collaboration

Competencies

  • AI Agents
  • Product Design
  • Beauty Tech
  • Enterprise

GenAI · Innovation

Forvia-X

Forvia · R&D Process Reinvention

Summary

Forvia's innovation and R&D teams wanted to use GenAI and needed to decide where to invest. I led a review of the process across departments, mapped uses for GenAI, assessed their value, and prioritized a roadmap for development.

Key outcomes

60+
use cases mapped
10
POCs built
8-pt
scoring matrix
4 mo
engagement

What I did

  • Identified pain points across all innovation teams and departments
  • Uncovered 60+ GenAI use cases spanning research, ideation, product development and handover
  • Assessed R&D pain points and overall GenAI maturity
  • Built an 8-point criteria matrix to score and prioritize use cases
  • Developed the POC-to-MVP roadmap and handover plan

Business impact

  • Prioritized 10 use cases from a list of 60+
  • 10 selected use cases built as POCs for global testing at Forvia
  • Gave leadership a GenAI investment plan grounded in the use-case assessments

Competencies

  • GenAI
  • Process Design
  • Value Case
  • Roadmap

AI project

SongAI GenAI platform

Accenture Song · Global Internal Platform

The challenge

Creative teams were experimenting with GenAI across Accenture Song, using separate tools and workflows. Project knowledge was hard to reuse. They needed a shared place to research, develop ideas, create content, and deliver work with AI.

The idea

SongAI is a modular GenAI platform for Accenture Song. It brings proprietary assets, third-party AI tools, and project knowledge into one place, with workflows that connect research, ideation, content creation, and delivery.

Key outcomes

70+
practitioner interviews across roles, practices, and markets
30+
GenAI opportunity areas mapped across the creative value chain
10
high-priority modules defined for the roadmap
3
Alpha modules launched in the global platform foundation
1
unified strategy replacing a fragmented GenAI tooling ecosystem

What I did

  • Led the product strategy and concept development from discovery through Alpha launch
  • Researched practitioner needs, identified opportunities, and defined the platform's purpose
  • Defined modules, prioritized the roadmap, and coordinated workflow design and launch activities
  • Built modular AI applications for research, content, ideation, testing, design, and delivery
  • Designed reusable workflows that connect AI modules into consistent ways of working
  • Designed the conversational interface for access to AI tools, project knowledge, and information
  • Designed a scalable architecture integrating internal capabilities and third-party AI tools

Business impact

  • Launched three Alpha modules as the foundation of the global platform
  • Brought separate GenAI tools and workflows into one platform
  • Gave creative teams a shared way to use AI from research through delivery
  • Connected secure AI modules to support human judgment in research, strategy, creative work, and delivery

Competencies

  • GenAI Strategy
  • Product Strategy
  • Platform Design
  • AI Workflow Design
  • Practitioner Research
  • Creative Operations
The Quest
Arabian Leopard
Roblox World

Metaverse · Gaming

Quest for Hope

RCU Saudi Arabia · AlUla Activation

Summary

Fewer than 200 Arabian leopards exist in the wild. The Royal Commission for AlUla wanted to engage people around the world in their conservation through a playable experience.

I developed Quest for Hope, a decentralized conservation campaign on Roblox and Decentraland. It launched on the first UN International Day of the Arabian Leopard (10 February 2024). Players explore digital recreations of AlUla's mountains, Old Town, and Hegra, where 18 leopard twins each represent a real animal being reintroduced into AlUla's wilderness. They photograph and collect the leopards and learn about biodiversity loss through reward-based activities in Valley Rush, Old Town Maze, and Cosmic Crossing. Rare leopard collectibles are in-game rewards and symbolic adoption links to the conservation programme run by RCU, Panthera, Catmosphere, and the Arabian Leopard Fund. The wider campaign included billboards in New York, London, Paris, and Beijing, a campaign film, and a 7km Community Catwalk Trail in AlUla.

Key outcomes

19.1M
press views
18
digital leopard twins
2
platforms
4
global billboards

What I did

  • Developed the campaign's creative concept
  • Designed the decentralized conservation mechanic linking digital leopard twins to real-world reintroduction tracking
  • Built the gaming experience across both Roblox and Decentraland platforms
  • Digitally recreated AlUla landmarks (Hegra, the Old Town, the mountain landscapes) as explorable in-game environments
  • Designed play-to-learn game mechanics: Valley Rush, Old Town Maze, Cosmic Crossing, leopard photography quests
  • Created the rare leopard collectible / NFT adoption system tying gameplay to the real conservation programme
  • Coordinated the campaign to launch on the first UN International Arabian Leopard Day (10 Feb 2024)

Business impact

  • 19.1M press coverage views globally
  • 20.9K social media views
  • 3K Roblox game visits at launch
  • Billboard activations across New York, London, Paris and Beijing
  • Launched as part of the global "Leap of Hope" campaign alongside a campaign film and 7km Community Catwalk in AlUla
  • Featured in Arab News, Fast Company Middle East, and global PR coverage
  • Brought AlUla's conservation mission to younger, gaming-native audiences worldwide

Competencies

  • Roblox
  • Conservation
  • Gaming
  • NFT

Visit live product Read case study