Florian Follonier

Profile

I am a Senior Cloud Solution Architect at Microsoft and a hands-on builder. My work sits where AI strategy has to become a real system: agentic applications, evaluation, observability, enterprise integration, and the engineering it takes to move from prototype to production. Before that I spent years across data platforms and machine learning consulting, which is where Relataly started. Outside customer work I build products and open-source projects, from source-linked AI research and ML-driven market tools to Swiss German speech benchmarks and an embodied-agent robotics simulator. Building things myself keeps my architecture advice honest.

Experience & education

1. Current

   Microsoft: applied AI and cloud architecture

   Working with customers and teams on Azure, Microsoft Foundry, agentic applications, evaluation, and the path from AI prototype to production.

2. Ongoing

   Independent products and open source

   Building and running public products and open-source projects: AIUseCaseHub, Rankoon, PerpXray, speech benchmarks, coding-agent evaluation, and robotics simulation.

3. Since consulting

   Public AI and ML publishing

   Writing practical articles and notebooks about machine learning, data science, analytics, and generative AI on Relataly, based on topics I explored as a consultant and builder.

4. Until 2018

   PhD, University of Bremen

   A research foundation that still shapes how I think about systems, adoption, innovation, and the gap between ideas and implementation.

Core capabilities

• Agentic systems and orchestration
• AI evaluation and observability
• Azure and Microsoft Foundry architecture
• Speech and multimodal AI
• Data platforms and ML engineering
• Workshops, writing, and technical storytelling

Agentic systems that can be operated

I help teams design agentic applications with clear boundaries: identity, tools, state, approvals, cost, and observability. The interesting work usually starts after the demo.

Evaluation before opinion

Benchmarks, groundedness checks, executable rewards. Whether it is speech models on Swiss German dialects or coding agents on real pull requests, I prefer decisions backed by measurements.

Products and experiments

I build small, concrete products to learn which assumptions survive real data and real users. Some stay live, some get stopped. Both teach me something I bring back into architecture work.

Selected projects

These publications appeared under my former name, Florian Müller.

1. 2018

   Patterns of Service Innovation: An Empirical Study of Automotive Engineering Services in Germany

   University of Bremen · Doctoral dissertation

2. 2017

   Service Innovation: Sensing with Information Systems

   Americas Conference on Information Systems (AMCIS) · Conference paper

3. 2016

   Schools of Thought in Service Modularity

   European Conference on Information Systems (ECIS) · Conference paper

Speaking & workshops

• What it takes to move from AI demos to systems people can use
• Agentic systems, orchestration, and where the hard parts start
• Evaluating agents and models with evidence instead of vibes
• MCP and tool ecosystems: making enterprise knowledge usable for agents
• AI governance, sovereignty, and architecture trade-offs
• How data platforms change when AI becomes part of everyday work

Formats

Keynotes · Conference talks · Hands-on workshops · Panels · Advisory sessions

Audiences

Leadership teams · Architects and engineering leads · Product and innovation teams · Data and AI practitioners · Public-sector and regulated organizations