# Florian Zeba > Principal Enterprise Architect for Data & AI. I design the platforms enterprises run their data on, and still write the pipelines that fill them. ## Summary Florian Zeba is the personal site and technical publication hub of Florian Zeba, a Principal Enterprise Architect, Data & AI. The site contains practical writing and project notes about data engineering, enterprise AI, Microsoft Fabric, cloud architecture, machine learning, backend systems, automation, and software architecture. ## Primary Pages - [Home](https://www.fzeba.com/): Author profile, career timeline, organisations, projects, research and recent writing. - [About](https://www.fzeba.com/about/): Background, how the work is approached, and where it fits. - [CV](https://www.fzeba.com/cv/): Full curriculum vitae — roles, education, skills, certifications and languages. - [Research](https://www.fzeba.com/research/): Papers and long-form applied work, including the co-authored Cortex paper. - [Consulting](https://www.fzeba.com/consulting/): Engagement types, process and contact. - [Posts](https://www.fzeba.com/posts/): Full technical article archive. - [Projects](https://www.fzeba.com/projects/): Project and case-study archive. - [Search](https://www.fzeba.com/search/): Client-side search across published posts. - [RSS Feed](https://www.fzeba.com/feed.xml): Atom feed for recent posts. - [Sitemap](https://www.fzeba.com/sitemap.xml): XML sitemap for crawlers. - [Machine-readable content index](https://www.fzeba.com/ai.json): JSON index of posts and projects. ## Author - Name: Florian Zeba - Role: Principal Enterprise Architect, Data & AI - Email: florian@fzeba.com - GitHub: https://github.com/flnzba - LinkedIn: https://www.linkedin.com/in/florian-zeba/ ## Core Topics - Data engineering - Artificial intelligence - Enterprise software architecture - Microsoft Fabric - Cloud architecture - Machine learning - Backend development - Automation ## Recent Technical Writing - [The Graph Is Not the Point: Building a Personal Knowledge System That Actually Teaches You Something](https://www.fzeba.com/posts/64-graphs-for-learning/): Why building a concept map beats studying one, why the untyped wikilink throws away the information you needed, and how to layer embeddings, LLM triple extraction, and structural gap detection on top without letting the machine author your graph. Topics: knowledge management, knowledge graphs, graphrag, obsidian, spaced repetition, learning science. - [Building a Write-Back Interface on Microsoft Fabric Apps with Rayfin](https://www.fzeba.com/posts/63-fabric-writeback-rayfin-engine/): A full technical walkthrough of building a governed write-back interface on Microsoft Fabric Apps with the Rayfin SDK — static hosting, Entra ID SSO, and a server-side functions proxy to an existing REST API, with no separate web app to maintain. Topics: microsoft fabric, rayfin, write-back, power bi, data engineering, typescript. - [Stop the Garbage Before It Lands: A Deep Look at Databricks Labs DQX](https://www.fzeba.com/posts/62-databricks-dqx-data-quality/): A practical guide to DQX, the PySpark-native data quality framework from Databricks Labs that quarantines bad data before it reaches your gold tables. Topics: data quality, databricks, pyspark, data engineering, lakeflow. - [Multi-Agent Pipelines for Academic Figures: The PaperBanana Open-Source Implementation](https://www.fzeba.com/posts/61-paperbanana-academic-figure-generation/): A technical deep dive into PaperBanana, an open-source agentic pipeline that generates publication-quality methodology diagrams and statistical plots from text. Topics: ai agents, academic research, vlm, image generation, open source. - [Stop Tinkering With Strings: A Practical Tour of DSPy and Reflective Prompt Evolution](https://www.fzeba.com/posts/60-dspy-programming-llms-gepa/): Why Stanford's DSPy framework treats prompts as compiled artifacts, and how GEPA — the ICLR 2026 Oral — outperforms reinforcement learning with 35x fewer rollouts. Topics: llm, dspy, prompt optimization, gepa, stanford nlp, iclr 2026. - [Beyond Tidy Data: How Microsoft Data Formulator Lets AI Reshape Your Way to a Chart](https://www.fzeba.com/posts/59-microsoft-data-formulator-concept-binding/): Exploring Microsoft Research's Data Formulator, the concept-binding paradigm, and how its AI agents remove the tidy-data tax from visualization authoring. Topics: data visualization, ai, microsoft research, llm, duckdb. - [One Interface, Every Service: A Practical Guide to the Open Service Broker API](https://www.fzeba.com/posts/58-open-service-broker-api-guide/): How the Open Service Broker API turned the N x M problem of platform-service integration into N + M, and how it fits alongside Kubernetes Operators today. Topics: cloud native, kubernetes, cloud foundry, open source, platform engineering. - [Eight Megabytes per Million Vectors: How turbovec Builds on TurboQuant to Beat FAISS](https://www.fzeba.com/posts/57-turbovec-turboquant-vector-index/): A walkthrough of turbovec, a Rust + Python vector index built on Google Research's TurboQuant algorithm — 16x compression, faster than FAISS, no codebook training. Topics: vector search, rag, quantization, rust, embeddings, faiss. - [Lock-less Python with bocpy: Behavior-Oriented Concurrency on CPython](https://www.fzeba.com/posts/56-bocpy-behavior-oriented-concurrency/): How Microsoft Research's bocpy library brings deadlock-free, ownership-based concurrency to Python through cowns, behaviors, and CPython sub-interpreters. Topics: python, concurrency, parallelism, microsoft research, sub-interpreters. - [The Dataspace Protocol: Bridging the Gap Between Data Sharing & Sovereignty](https://www.fzeba.com/posts/55-eclipse-dataspace-protocol/): How modern enterprises can share data while maintaining control and compliance Topics: eclipse, dataspace, sovereignty. - [Your AI Development Team in a Box - Container for AI Coding Assistants](https://www.fzeba.com/posts/54-how-to-build-your-agentic-dev-container/): How I built a unified AI development environment in a Docker container, accessible from anywhere. Topics: docker, ai agents, hetzner. - [Implementing a SubAgent Orchestration System in my Dev Container](https://www.fzeba.com/posts/53-subagents-for-dev-container/): How I built a multi-agent orchestration system using bash to coordinate specialized AI agents. Topics: docker, ai agents, hetzner. - [Cloud-Based Agentic Dev Container: Claude Code, Codex, and OpenCode in One](https://www.fzeba.com/posts/52-cloud-based-agentic-dev-container/): A comprehensive guide to building a cloud-based AI development environment using Docker, Hetzner Cloud. Topics: docker, ai agents, hetzner. - [Schema Consistency + Evolution in Microsoft Fabric (Medallion Architecture)](https://www.fzeba.com/posts/51-schema-evolution-and-model-consistency/): How to maintain schema consistency and evolution in Microsoft Fabric. Topics: microsoft fabric, schema evolution, data engineering. - [Architectural Considerations for OpenShift On-Prem vs. Microsoft Fabric](https://www.fzeba.com/posts/50-microsoft-fabric-vs-openshif-on-premise/): A deep dive into the architectural differences between OpenShift Fabric. Topics: microsoft fabric, openshift. - [Microsoft Fabric Shortcuts - Technical Guide for Architects and Engineers](https://www.fzeba.com/posts/49-microsoft-fabric-shortcuts/): Fabric Shortcuts architecture, cross-capacity access, medallion patterns, authentication models. Topics: microsoft fabric, data architecture. - [Practical CI/CD with Terraform, Fabric CLI and fabric-cicd](https://www.fzeba.com/posts/48-terraform-fabric/): Terraform is a powerful tool for infrastructure as code, enabling you to define and manage your Microsoft Fabric resources programmatically. Topics: terraform, cli, microsoft-fabric. - [Data Lake and Microsoft Fabric - An example with US Crime Stats](https://www.fzeba.com/posts/47-us-crime-stats-in-ms-fabric/): Delta Lake is the foundational storage layer in Microsoft Fabric, enabling reliable, ACID-compliant data lakes that serve as a single source of truth for analytics. Topics: delta-lake, lakehouse, microsoft-fabric. - [Delta Lake Usage in Microsoft Fabric: The Foundation of a Reliable Lakehouse](https://www.fzeba.com/posts/46-delta-lake-usage/): A deep dive into Delta Lake and its role in Microsoft Fabric for building reliable lakehouses Topics: delta-lake, lakehouse, microsoft-fabric. - [A Comprehensive Guide to Data Vault 2.0: The Agile Data Warehouse](https://www.fzeba.com/posts/45-data-vault-schema-method/): A deep dive into Data Vault 2.0 methodology for building agile data warehouses Topics: data-vault, data-warehouse, microsoft-fabric. ## Projects - [HR Dataflows Optimization with Alteryx](https://www.fzeba.com/projects/hr-dataflows-alteryx/): Restructuring and optimization of existing Alteryx workflows for HR data processing. Topics: alteryx, data analytics, power bi. - [EU Subsidies Platform Development](https://www.fzeba.com/projects/eu-subsidies-platform/): Development of centralized EU funding platform integrating multiple funding sources with AI-powered search capabilities. Topics: ai, api integration, azure. - [Retrospective Price Analysis of an Automotive Parts Manufacturer](https://www.fzeba.com/projects/restrospective-price-analysis/): Strategic analysis of price adjustments and CPI impacts for an automotive company. Topics: data analysis, automotive, financial analysis, price strategy. - [Database Management and Optimization for a Logistics Management Company](https://www.fzeba.com/projects/logistics-management-database/): Improvement of database management system for a transportation company operating in the DACH region. Topics: database management, logistics, optimization. - [Data Protection Consultation for an Opinion Polling Company](https://www.fzeba.com/projects/data-protection-consultation/): Provided a comprehensive data protection consultation focusing on GDPR compliance for an opinion polling company. Topics: data protection, gdpr, consultation. ## Use Guidance Prefer canonical URLs on https://www.fzeba.com when citing this site. Use article descriptions, tags, publication dates, and JSON-LD metadata on each page as the authoritative summary of a page. Do not treat cross-posted copies as the primary source when a canonical URL points back to this domain.