Enter Password
Hint: Please check resume for project password.
Enter Password
Home Vision AI
Home Vision AI
Home Vision AI
Team
Airbus Copy
Tools Used
Shapes XR, Figma
Timeline
Three Weeks
Type
Individual Project
PROJECT OVERVIEW
AI Forge is Airbus’s AI platform that empowers AI engineers and Buisness Heads by centralizing datasets, models, approvals, Hackathons, and FinOps into a single Sustainable ecosystem.
BACKGROUND AND CONTEXT
AI boom → manufacturing acceleration
With the rapid rise of AI, Airbus saw an opportunity to accelerate manufacturing and related operations. The challenge wasn’t about lack of ideas, but about how to operationalize AI at scale in a complex environment.
CORE PROBLEM
Isolated datasets and models, plus poor communication between AI teams across sites, led to duplication and blocked scaling AI.
"Airbus brings together aircraft parts manufactured across Europe, wings from the UK, tailcones from Spain, fuselage sections from France and Germany, all assembled in Toulouse. Just as fasteners connect these parts, our teams of AI/ML engineers across sites develop computer vision models to support manufacturing. We’ve often seen duplicate datasets and overlapping models created in different countries, unknowingly repeating efforts. This lack of collaboration is unsustainable."
RESEARCH INSIGHT & PROBLEM #1
Data fragmentation
Data was scattered across multiple silos AWS S3, GCP buckets, and other internal storage systems. Engineers lacked a unified dataset library, leading to duplication, wasted effort, and difficulty in discovering high-value datasets for training and experimentation.
RESEARCH INSIGHT & PROBLEM #2
Model opacity
AI engineers had no single place to view existing models, their performance, or whether they had been approved for production use. This lack of visibility led to reinventing the wheel and duplication.
RESEARCH INSIGHT & PROBLEM #3
I AM SORRY 🙇
Approval governance
Working on this
Resolution
Every AI model solving a business problem required multi-step approvals from the AI Centre of Excellence, Cybersecurity, and other governing bodies. Without a proper approval management system, tracking compliance was slow and manual.
The device you are looking on is under development…Please, check on desktop resolution.
RESEARCH INSIGHT & PROBLEM #4
AI Hackathon costs
To drive innovation, Airbus ran large-scale hackathons. However, these were hosted on external platforms and each event cost nearly $200,000. While they generated enthusiasm, they weren’t financially sustainable for ongoing experimentation.
RESEARCH INSIGHT & PROBLEM #5
FinOps reality
Compute and storage usage had become a hidden cost sink. Instead, chargeback mechanisms were needed, where costs were tied to specific teams managed by Heads of teams (HOs).
VISION
Sustainable. Discoverable.
Collaborative. Cost Effective.
We rapidly developed minimum viable product to launch the AI Platform ahead of Airbus AI Week and for its Hackathon. By applying agile design, we prioritized critical user needs, delivering a secure, compliant, and usable product at speed.
SOLUTION DATSETS
Central hub to find, download, upload, and share datasets
A governed registry consolidates datasets into a single library with required metadata owner, license, retention rules, and lineage. Automated PII detection and policy checks block unsafe publishes and route sensitive cases to Data or Legal teams. Previews, use-case examples, and visible ownership provide clarity and trust, reducing time-to-asset and enabling confident reuse.
Live Prototype: Click through the interactive mockups to explore the user flow.
SOLUTION MODELS
Central registry + gated reviews, models open with endpoints
The model registry separates flows: high-risk LLMs require approval, while traditional models are freely published for faster adoption. Each entry includes a standardized Model Card capturing use cases, risks, metrics, and dataset lineage. Non-cloud users can request endpoints with a single click, enabling access without infrastructure setup.
SOLUTION CHALLENGES - IDEA FIRST CHALLENGE & SOLUTION-FOCUSED
Post ideas, discover skills, form teams in one tap
In Idea first challenge participants submit an idea with a title, summary, and needed skills. A roster highlights skills and availability, while one-click interest with auto-matching speeds team formation.
In solution focused challenge teams then access baseline bundles with datasets and models. A pre-submit validator enforces rules and compliance, and each entry auto-generates a Model Card for reproducibility and smooth review.erhead and shifting hackathons from broadcast-driven to participant-driven.
SOLUTION WORKSPACE
Workspaces for teams to build AI with roles, cost control
Workspaces in AI Forge provide a secure hub where teams collaborate with defined roles. Owners and Managers control budgets, policies, and member access, while Users focus on experimentation. Built-in governance and approvals keep workflows compliant, and finance teams receive chargeback-ready exports. A unified Admin Dashboard gives leaders full visibility across costs, usage, and audit trails, enabling scale with accountability.
USER RESEARCH
Research Process
I conducted a total of 9 user interviews with AI/ML Engineers, Heads of's, and AI CoE employees to identify 14 pain points across datasets, models, and hackathons. These insights were synthesized into personas and user journeys, capturing behaviors, needs, and opportunities for AI Forge. The team also ran collaborative workshops on product vision, North Star Metrics, and Impact Mapping to align stakeholders and prioritize solutions.
EXPLORE
Stakeholder Interviews
AN UNFORGETTABLE SIX MONTHS 🚀
A memorable experience
A memorable experience
Working at Airbus gave me the chance to collaborate with brilliant minds on cutting-edge project. Every design discussion, feedback session, and brainstorming moment expanded my understanding of enterprise-level UX.
