SAP d-com · Mannheim
Situation-aware, agentic support.
Co-presented with Alexander Gahr on bringing product context and guided resolution directly into the support experience.
A line-mesh study morphing Munich's Frauenkirche into Casablanca's Hassan II Mosque.
Full-stack software engineer at SAP
I build software end to end.
I work across product, platform, and data: turning complex systems into software people can understand and trust. Today I build Built-In Support at SAP; my recent research explores generative super-resolution for satellite-based solar mapping.
Selected highlights
Currently
Building Built-In Support at SAPEducation
MSc Data Engineering & Analytics, TUM '25Research
Published by IEEE · ACDSA 2026Selected work
Two projects in depth.
What I am building at SAP and what I researched for my master's thesis.
Enterprise platform / SAP
Building support into the product experience
End-to-end feature ownership for a large-scale, microservice-based support platform.
Context
Built-In Support brings help and support closer to where customers work, across a complex enterprise product landscape.
My contribution
I take features from requirements and epic planning through secure-by-design implementation, including threat modeling and cross-team technical decisions.
Outcome
The focus is durable platform work: improving reliability and maintainability while making new capabilities easier to evolve.
Applied AI / TUM research
Sharper satellite imagery for better solar mapping
A modular research pipeline evaluating whether generative super-resolution can improve photovoltaic-panel segmentation.
Context
Accessible satellite imagery has broad coverage, but its spatial resolution makes small photovoltaic installations difficult to detect reliably.
My contribution
I assembled the workflow from geospatial acquisition through ESRGAN, S2DR3, and Satlas enhancement, image-quality evaluation, and an interactive segmentation interface.
Outcome
The study found that super-resolution improved segmentation, while perceptual image quality did not consistently predict detection performance. It was presented orally at ACDSA 2026 and published by IEEE.
More work
Other projects.
DelayBahn
A distributed journey planner that combined European rail routing with Deutsche Bahn delay prediction.
Processed more than 2 million trips through scheduled ingestion pipelines.
Plato 2.0
A full-stack platform for semantic search, similarity detection, and clustering across large image collections.
Designed for datasets ranging from 100,000 to more than 1 million images.
GenAI for crop breeding
Refactored a data-driven AI platform and added conversational data exploration, visualization, and pipeline construction.
Geometry: Bavaria LoD2 + © OpenStreetMap contributors + Hassan II reference by abdlilah ben (CC BY 4.0)
About me
Salam, I'm Ziyad.
I grew up in Morocco, studied computer science at Al Akhawayn, taught database systems, and moved to Germany for an MSc in Data Engineering & Analytics at TUM.
I care about collective wins, direct communication, and software that earns trust. Outside work, you'll find me following Liverpool and the Atlas Lions, playing chess, or arguing about hip-hop.
More about meCurrently
What I'm up to.
At work
Owning full-stack features for SAP Built-In Support, with a focus on secure implementation and maintainable platform foundations.
In research
Sharing the IEEE-published photovoltaic super-resolution work presented at ACDSA 2026 and keeping its research tooling open.
Beyond the screen
Following Liverpool and the Atlas Lions, playing chess, FIFA, and Smash Bros, and keeping hip-hop in the rotation.
Want to work together?