Freelance Principal Technology Consultant

Azure Modernization
Partner Connectivity
Integration Security
GenAI Engineering

Transavia, the Dutch airline within the Air France-KLM group, engaged Kelderro for four fixed-term assignments, each with its own scope, deliverables and end date: two legacy AS-400 migrations to Azure, partner impact for a core reservation platform upgrade, and a secure partner gateway.

Schiphol, The Netherlands
Freelancer
  1. Apr 2023 – Dec 2023 Allowance payment system
  2. Jan 2024 – Apr 2024 Flight crew review election system
  3. May 2024 – 2025 After-booking & reservation platform upgrade
  4. 2026 Partner gateway
fixed-term assignments
4
partner outages during platform upgrade
0
saved per year
€48k
legacy AS-400 systems moved to Azure
2

Key Impact

  • Moved two critical systems off the AS-400 to Azure serverless before Transavia's last AS-400 programmer retired: crew compensation became a predictable, hands-off process, and replacing the crew review election tool enabled AS-400 decommissioning, saving €48k a year in mainframe run costs
  • Assessed the partner impact of a core reservation platform upgrade and adapted the integration endpoints: breaking changes caught before go-live, zero partner outages
  • Designed and built a partner gateway that limits each partner to its allowed endpoints, with in-depth request and response filtering and masking, moving the logic from API Management policies into a tested .NET service
  • Delivered the gateway with a coding agent and presented the approach at Transavia's AI Guild, after which two tech leads requested a knowledge-sharing session so their own teams could adopt GenAI-assisted engineering
The engagement

Transavia has engaged Kelderro for four fixed-term assignments since 2023, each with its own scope, deliverables and end date. It started with a single legacy migration, and each follow-up assignment was awarded on the strength of the one before.

2023 · Allowance payment system

April 2023 - December 2023

Challenge. Crew allowances were calculated and paid from an IBM AS-400 system. It was expensive to run and relied on manual work to keep payments correct. More urgently, the knowledge to maintain it was disappearing: the last person at Transavia who could program the AS-400 was about to retire.

Approach. Rob gathered the requirements and worked closely with the subject-matter experts to understand which payments were being corrected by hand, and why. Together they reached agreement on rules that could replace those manual corrections, and Rob designed the solution that automates them. He owned the rest of the lifecycle too: architecture, development, infrastructure and post-delivery support, rebuilding the system on Azure serverless and integrating it with Transavia’s existing systems.

Outcome. Allowances are now calculated by the agreed rules instead of corrected by hand. That means fewer payment errors, a faster payroll cycle, and a clear trail showing how each crew member’s allowance was calculated. Management gained insight into allowances through Power BI, and around one FTE of manual correction work was freed up. This first assignment also earned the ones that followed.

2024 · Flight crew review election system

January 2024 - April 2024

Challenge. A second critical AS-400 system decides which flight crew review, or are reviewed, during flights, based on predefined criteria. It had to move off the legacy platform before that last AS-400 knowledge left the company, without interrupting operations.

Approach. Rob took the same end-to-end ownership and rebuilt the tool as an event-driven Azure serverless solution (Event Grid, Functions, Storage). He worked closely with the maintainer of the old system, who until then had to intervene by hand, with subject-matter experts and with the cabin lead, until all of them confirmed that the new system applied the election rules correctly.

Outcome. A secure, maintainable Azure solution, handed over with technical support and onboarding. With this last workload moved off, the AS-400 could be decommissioned, saving Transavia €48k a year in mainframe run costs.

2024 - 2025 · After-booking & reservation platform upgrade

May 2024 - 2025

Challenge. Transavia upgraded its core reservation platform, the system every partner booking runs through. Changes in its behaviour could break the API and B2B integrations that partners use to sell and change tickets after booking.

Approach. Rob ran the partner impact analysis, mapping which changes would affect partner integrations. Where possible, he adapted Transavia’s own integration endpoints so partners were shielded from the changes. Where partners did need to act, he briefed them and the business on what would change and when.

Outcome. Breaking changes were caught and fixed before go-live, and partners kept selling and changing tickets through the upgrade with no outages.

2026 · Partner gateway

2026

Challenge. Large travel integrators connect directly to Transavia’s reservation platform. Its API exposes far more operations and data than any single partner needs, and every extra field is data that can leak.

Approach. Rob designed a smart wrapper in front of the reservation platform that limits each partner to the endpoints it is allowed to use. It filters and masks requests and responses in depth, so only the minimum information crosses the boundary. The first attempt used Azure API Management policies alone, but complex filtering logic in policies proved inflexible and hard to test. Rob moved that logic into a dedicated .NET service: partners call API Management, which handles the gateway concerns, and the .NET service applies the filtering and masking before anything reaches the reservation platform.

GenAI in practice. Rob planned and built the service with a coding agent, using Matt Pocock’s Wayfinder skill to plan work that ran over many sessions, where the route wasn’t clear up front. He then presented the result to other teams and at Transavia’s internal AI Guild, including the timeline and the steps they would need to take themselves. Showing a working example rather than a slide deck convinced two tech leads, who requested a knowledge-sharing session and some ad-hoc advice so their own teams could adopt the approach. A senior architect also took an interest in the material and asked to learn more.

Outcome. Partners get a narrow, purpose-built API instead of the full platform. The filtering and masking rules live in tested code that can be changed with confidence.

Technical decisions
  • Durable Functions for allowance calculations. Calculating and paying crew allowances is a long-running, multi-step batch. Durable Functions orchestrates it reliably on Azure Functions, without servers to run.
  • The data store follows the data. Each system got the store that fit its data shape: Azure SQL Database for one, Azure Cosmos DB for another. Where a database would have been overkill, simpler storage did the job: Blob Storage for exchanging information between systems, and Table Storage for tracking system state and keeping an audit trail of every run.
  • .NET Aspire for the developer loop. Aspire gives one-command local environments and built-in observability. Deployment to Azure stays on the regular pipelines.
  • Code over gateway policies for complex logic. API Management stays the front door for partner traffic, but request and response filtering moved into a .NET service where it can be unit-tested, versioned and evolved.
  • No new single point of knowledge. After an AS-400 that only one person could still maintain, the replacements were built on mainstream .NET and Azure services and kept on supported versions, so Transavia’s own teams can maintain them.