The Project

LEAD:FUH is a research project developing a university-wide data architecture for Learning Analytics. Think of it as a backbone that connects Moodle, SAP, and other university systems, harmonizes their raw data, and produces research-grade datasets alongside real-time insights for students, faculty, and advisors.

The use cases range from early dropout detection and personalized learning recommendations to self-monitoring dashboards where students can track their own progress. I created the initial version of the interactive platform visualization on that page, and together with the website team we brought it to life.

My role in the project: software architect with a focus on the Moodle platform, system architecture, data flow management, and hosting of the Learning Analytics infrastructure.

The Tech

If you've been following the Modern Data Stack conversation, this will look familiar, but with a twist. We're running ClickHouse for behavioral logs, PostgreSQL for consent management and results, dbt for transformation, FastAPI as a serving layer with live consent checking, DataHub for governance and lineage, and Dagster for orchestration. PeerDB handles CDC for real-time ingestion, Meltano covers batch ELT.

The twist is the data. This isn't clickstream data for conversion optimization. This is highly sensitive student data: grades, learning behavior, personal information. Every single API request runs through a consent gateway. No opt-in, no data. Revocation triggers immediate blocking and physical deletion. Privacy isn't a feature here; it's the foundation.

For someone who spent years building commerce platforms with >99.9% uptime, bringing that same rigor to a university research project is both humbling and deeply satisfying. The stakes are different, but the architectural discipline is the same.

Why I Made the Switch

Two reasons, and I'm going to be honest about both.

First: meaning. After twenty years of building things that help companies sell more stuff, I wanted to use my expertise for something with more direct societal impact. Learning Analytics has the potential to genuinely help students, especially at a distance university like FernUniversität, where dropout rates are a real problem and personalized support can make a tangible difference. Building the infrastructure that enables this feels like the right use of what I've learned.

Second: work-life balance. I've written about burnout before. I've experienced it. The startup-to-acquisition-to-scale cycle at Frontastic and commercetools was incredible, but it came at a cost. I have a daughter. I have a garden. I like to cook. I wanted a role where I could do meaningful technical work without sacrificing the parts of life that keep me sane.

Both reasons turned out to be exactly right.

What Surprised Me

The team. I didn't expect this to be one of the best teams I've ever worked with. It's incredibly diverse, researchers, data engineers, educators, student assistants. People from completely different backgrounds working toward the same goal. The spirit is real, not corporate-poster real, but genuinely collaborative in a way I rarely experienced in the commercial world.

The challenges. Building a Modern Data Stack for a university is not a downgrade in complexity. If anything, the constraints are harder. Consent management at the data layer is a non-trivial architectural problem. Bringing Business Continuity Management to a research project that wasn't designed for it requires careful negotiation between academic culture and operational reality. Getting security right for highly sensitive data while keeping the system accessible for researchers, that's the kind of puzzle I love.

The learning. After years of leading teams, I'm back to being the person who also writes code. I'm getting deep into Moodle internals, learning about data mesh principles applied to educational data, and working with tools I'd only read about. It's refreshing. It reminds me why I started coding in the first place.

The Uncomfortable Part

Let me be transparent: the pay is significantly lower than what I earned in the commercial tech world. Universities operate on public funding, and salaries reflect that. This was a conscious trade-off, and it only works because of my financial position after years in the industry.

I'm aware that not everyone has this privilege. But for those who do — especially senior engineers and CTOs who've had a good run commercially — I'd encourage you to think about it. The impact-per-hour ratio in projects like this is extraordinary. Your skills are rare in the academic world, and the problems are genuinely interesting.

What Comes Next

LEAD:FUH is designed for sustainability. The plan is to hand over operations to the university's IT center (ZDI) after the project phase, while the team continues maintaining ML models and data products. Infrastructure as Code means the whole thing can be rebuilt from backups. That's the kind of thinking I brought from the commercial world, and it fits perfectly here.

I don't know if this is my "forever" role. But right now, it's exactly where I want to be: doing architecture work that matters, in a team I enjoy, with time for the things outside of work that make life worth living.