Skip to content
ArnaudDeveugle
AI Venture Builder & Product Strategist

Strategy meets
Autonomous AI.

Bridging the gap between high-level business strategy and hands-on autonomous AI execution.

View Work ↗
Experience
14+ yrs
Based in
Ghent, Belgium

AI Projects

01Build time: 1.5 months

Support Platform

Used daily by 10+ external call center agents to resolve urgent incoming support cases, without ever touching core systems. When a customer calls, all the context of that user is displayed and quick actions are visible for fast resolution. A separate backend proxy isolates all database access and API calls, enforcing strict data isolation. The platform is fully multi-language and includes a fully isolated staging environment to train new agents safely, without any risk to live systems.

Claude CodeReactFastAPIAPI ProxyMulti-languageStaging Environment
02Build time: 1 month

Visual Workflow Builder

With 30+ custom workflows live in production, this internal builder standardizes complex support processes across the organization via drag-and-drop. Custom node types reflect real business operations, making domain-specific workflow creation fast and accessible. Workflows can also be auto-generated from plain text using AI. Every workflow exports as structured JSON, ready to be consumed by AI agents, making it a foundation for future automation. Replaced hours of manual script lookup with a structured, auditable decision layer.

Google AntigravityReactReact FlowPrompt-to-Workflow
03Build time: 3 weeks

Self-Compiling Company Wiki

Eliminates manual documentation entirely by autonomously compiling company knowledge from raw sources (PDFs, Markdown, JSON). A 6-step compiler pipeline summarizes sources, extracts entities, builds a typed knowledge graph, writes interconnected articles, and links everything with cross-references. Users can flag errors or gaps in plain language via a snippet system; the AI finds the right article and updates it automatically. The wiki is queryable via dedicated APIs with persona-based responses, serving tailored answers to internal employees, customers, or any defined role. Multi-language, with hybrid AI search and full version history.

Claude CodeKnowledge GraphHybrid AI SearchPersona APIMulti-languageRAG
04Build time: 1 day

Custom AI Wiki

Built in a single day, this personal knowledge base has compiled 136 wiki pages from 25+ ingestions across 7 weeks. Dedicated skills handle 6 content types end-to-end: GitHub repos, YouTube videos, web articles, X posts, PDFs, and HTML. Unlike RAG, it uses an LLM-as-compiler paradigm: raw sources stay immutable while Claude distills them into structured pages, a navigable knowledge graph, and cross-source synthesis analyses. A three-tier retrieval system resolves most queries without opening more than 5 files.

Claude CodeLLM CompilerKnowledge GraphObsidianClaude Code Hooks
05Build time: 2 days

This Website

A professional CV website built in 2 days using AI-assisted development. Features smooth animations, a dual dark/light theme, and a responsive layout built to production standards, including a full test suite and accessibility support. A hands-on demonstration that the skills described on this page translate directly into shipped product.

Claude CodeNext.js 16TypeScriptTailwind CSS v4Framer MotionJest / RTL

Experience & Education

Experience

Battmobility

Head of Strategy & Innovation
February 2025 - Present · Ghent, BE

I joined Battmobility to align our technical infrastructure with our long-term product strategy, focusing heavily on optimizing our data layer and core architecture. Recognizing that immediate business value lay in operational efficiency, I pivoted from high-level roadmaps to hands-on execution using advanced AI tooling. I engineered a secure, full-stack support application that aggregates complex real-time context (bookings, lease details, and history) for an external call center. To protect our core architecture, I built a custom backend proxy that safely routes and isolates external database access. To date, I have shipped 5 new AI-powered internal applications that automate repetitive operational bottlenecks, while architecting autonomous AI agents for future-proof customer support.

5 AI apps shipped to productionCustom LLM proxy & secure routingAgent architecture for support
Experience

Karamel

Founder & CEO
June 2019 - May 2024 · Ghent, BE

As the founder and acting Product Manager of Karamel, my mission was to solve the systemic trust and verification issues within the HR-tech market. I spearheaded the data architecture to launch the world's first commercial platform utilizing decentralized 'SOLID' data pod technology, establishing a direct R&D partnership with Ghent University. Balancing the roles of vision and grit, I owned the end-to-end UX/UI strategy, designed a market-defining skill-based career data set, and managed a fully remote engineering team. Beyond product building, I mastered high-stakes venture pitching and engineered advanced financial cash-flow models that became internal benchmarks for operational scaling. Ultimately, me and my team grew the platform's ecosystem to over 800 companies and 5,000 active users across Flanders.

800+ companies onboarded5,000+ active usersWorld's first commercial SOLID platform
Experience

Securitas

Solution Manager
June 2016 - November 2017 · Brussels, BE

Following Securitas' strategic acquisition of the tech company Zenitel, I was brought in to bridge the gap between traditional physical security services and next-generation technical solutions. My core challenge was finding untapped market opportunities where technology could elevate human capital. During this time, I assisted in the launch initiative for Securitas' first autonomous tethered drone surveillance system, designing new value propositions for industrial site security and crowd control. This role gave me invaluable experience navigating complex corporate structures and deploying frontier hardware-software integrations.

Strategic new market discoverySolution management
Experience

Dstny

Product Manager
October 2013 - June 2016 · Brussels, BE

Stepping in as the very first Product Manager in a hyper-growth SaaS scale-up, my mission was to bring structure by building our product governance frameworks from scratch. I quickly noticed that our highly customizable, 'building-block' enterprise software was creating massive friction and slowing down adoption in the SME market. To fix this, I engineered a standardized, off-the-shelf SaaS package tailored for smaller businesses, complete with custom pricing strategy, technical documentation, and market positioning. To ensure scalability, I designed a structured, frictionless handoff process between our sales and technical delivery teams, and successfully initiated and negotiated a high-velocity white-label wholesale partnership model that opened entirely new indirect revenue channels.

First PM hire in a hyper-growth scale-upOff-the-shelf SaaS package for SMEsWhite-label wholesale partnerships
Education

Ghent University

Master of Science (ir.) Industrial Engineering and Operations Research
September 2010 - June 2012 · Ghent, BE

Skills

AI Execution
How I ship with AI today
Agentic WorkflowsPrompt EngineeringRetrieval-Augmented Generation (RAG)AI Opportunity AssessmentRapid PrototypingTest-Driven Development
Tech Stack
What I reach for to build
Claude Code / Anthropic APIReact / Next.jsTypeScriptTailwind CSSFramer MotionNode.js / PythonFastAPIVector DatabasesAPI Proxy & Security RoutingJest / RTLn8nZapier
AI Transformation
Shaping AI inside the business
AI Strategy & Roadmap DesignChange ManagementBusiness Model EvolutionAI Governance & PrivacyCross-functional LeadershipStakeholder ManagementAI Best Practices
Venture Architecture
My foundation as a builder
Product Management (SaaS)Analytical Thinking (ir.)Operations ResearchUX/UI Product StrategyFundraising & Investor Pitching
Dutch
Native
English
Fluent
French
Average
German
Average

Working with AI agents

Illustration of Arnaud at a laptop, conversing with AI agents shown as translucent glass formsIllustration of Arnaud at a laptop, conversing with AI agents shown as translucent glass forms
Arnaud Deveugle
AI Venture Builder & Product Strategist
ArnaudDeveugle
arnaud@aforsteri.com·+32 476 43 96 28·arnauddeveugle.com·linkedin.com/in/arnauddeveugle·23 October 1989·Merelbeke, Belgium
Profile

AI Venture Builder and Product Strategist who turns business strategy into shipped software. After 5+ years founding and running an HR Tech startup, I now build production AI systems hands-on, from autonomous agents to full-stack internal tools, pairing deep operational experience with rapid, AI-assisted delivery.

Expertise
AI Execution
Agentic WorkflowsPrompt EngineeringRetrieval-Augmented Generation (RAG)AI Opportunity AssessmentRapid PrototypingTest-Driven Development
Tech Stack
Claude Code / Anthropic APIReact / Next.jsTypeScriptTailwind CSSFramer MotionNode.js / PythonFastAPIVector DatabasesAPI Proxy & Security RoutingJest / RTLn8nZapier
AI Transformation
AI Strategy & Roadmap DesignChange ManagementBusiness Model EvolutionAI Governance & PrivacyCross-functional LeadershipStakeholder ManagementAI Best Practices
Venture Architecture
Product Management (SaaS)Analytical Thinking (ir.)Operations ResearchUX/UI Product StrategyFundraising & Investor Pitching
Languages
  • DutchNative
  • EnglishFluent
  • FrenchAverage
  • GermanAverage
Experience
Battmobility & MobilityOS · Ghent, BE2025 - Present
Head of Strategy & Innovation
  • Re-engineering the core architecture for the fleet management platform, aligning infrastructure with product strategy.
  • Developed a secure full-stack support app and a custom backend proxy for external call centers, ensuring isolated data access.
  • Deployed multiple high-impact internal AI applications to automate operational bottlenecks and repetitive workflows.
Karamel · Ghent, BE2019 - 2024
Founder & CEO (acting Product Manager)
  • Founded the HR-tech start-up, while acting as lead PM for the end-to-end UX/UI strategy and scaling to 800+ companies and 5k+ users.
  • Spearheaded the data architecture to launch the world's first commercial platform utilizing decentralized SOLID data pods.
  • Architected advanced financial cash-flow models and secured top spots at high-profile startup investment pitching events.
Securitas · Brussels, BE2016 - 2017
Solution Manager (M&A Integration)
  • Created joint B2B tech-enabled value propositions following the strategic acquisition of tech company Zenitel.
  • Assisted in the innovative launch initiative of Securitas’ first autonomous tethered drone surveillance system.
Dstny · Brussels, BE2013 - 2016
Product Manager
  • First Product Manager in a hyper-growth SaaS scale-up; designed core product governance frameworks from scratch.
  • Created a highly successful, standardized off-the-shelf SaaS package for SMEs, owning pricing and GTM execution.
  • Established a white-label wholesale partnership and optimized the internal sales-to-delivery handoff workflows.
Education
Ghent University
Master of Science (ir.) Industrial Engineering and Operations Research
September 2010 - June 2012 · Ghent, BE
ArnaudDeveuglePortfolio

Five products shipped with AI-assisted development in 2026 so far.

01Support Platform
Build time1.5 months

Used daily by 10+ external call center agents to resolve urgent incoming support cases, without ever touching core systems. When a customer calls, all the context of that user is displayed and quick actions are visible for fast resolution. A separate backend proxy isolates all database access and API calls, enforcing strict data isolation. The platform is fully multi-language and includes a fully isolated staging environment to train new agents safely, without any risk to live systems.

Claude CodeReactFastAPIAPI ProxyMulti-languageStaging Environment
02Visual Workflow Builder
Build time1 month

With 30+ custom workflows live in production, this internal builder standardizes complex support processes across the organization via drag-and-drop. Custom node types reflect real business operations, making domain-specific workflow creation fast and accessible. Workflows can also be auto-generated from plain text using AI. Every workflow exports as structured JSON, ready to be consumed by AI agents, making it a foundation for future automation. Replaced hours of manual script lookup with a structured, auditable decision layer.

Google AntigravityReactReact FlowPrompt-to-Workflow
03Self-Compiling Company Wiki
Build time3 weeks

Eliminates manual documentation entirely by autonomously compiling company knowledge from raw sources (PDFs, Markdown, JSON). A 6-step compiler pipeline summarizes sources, extracts entities, builds a typed knowledge graph, writes interconnected articles, and links everything with cross-references. Users can flag errors or gaps in plain language via a snippet system; the AI finds the right article and updates it automatically. The wiki is queryable via dedicated APIs with persona-based responses, serving tailored answers to internal employees, customers, or any defined role. Multi-language, with hybrid AI search and full version history.

Claude CodeKnowledge GraphHybrid AI SearchPersona APIMulti-languageRAG
04Custom AI Wiki
Build time1 day

Built in a single day, this personal knowledge base has compiled 136 wiki pages from 25+ ingestions across 7 weeks. Dedicated skills handle 6 content types end-to-end: GitHub repos, YouTube videos, web articles, X posts, PDFs, and HTML. Unlike RAG, it uses an LLM-as-compiler paradigm: raw sources stay immutable while Claude distills them into structured pages, a navigable knowledge graph, and cross-source synthesis analyses. A three-tier retrieval system resolves most queries without opening more than 5 files.

Claude CodeLLM CompilerKnowledge GraphObsidianClaude Code Hooks
05My personal website
Build time2 days

A professional CV website built in 2 days using AI-assisted development. Features smooth animations, a dual dark/light theme, and a responsive layout built to production standards, including a full test suite and accessibility support. A hands-on demonstration that the skills described on this page translate directly into shipped product.

Claude CodeNext.js 16TypeScriptTailwind CSS v4Framer MotionJest / RTL