Machine Learning Engineer @ Flower Labs.
Hello, nice to meet you. I build privacy-first machine learning systems, taking ideas from research to production with rigorous evaluation and dependable deployment.
About Me
For the past several years, I’ve had the pleasure of working alongside extremely talented engineers and researchers, developing a leading Federated Learning framework that brings cutting-edge, on-device machine learning solutions to life.
I’m passionate about on-device machine learning, with my work centered on cutting-edge projects that push the boundaries of technology. I’m particularly focused on developing innovative ML solutions that can be seamlessly deployed directly on devices.
As a machine learning engineer at Flower Labs, I'm developing Flower, a leading Federated Learning framework. My focus is on optimizing FL solutions for on-device deployment, and I’m dedicated to building software with intuitive design and technical excellence under the hood.
Outside my professional work, I enjoy developing aesthetically pleasing websites with excellent user experience. I'm keen on blending functionality with design to create intuitive and visually engaging interfaces. My personal web development projects are Farmily and Traveldigga.
I hold both an M.Sc. and a B.Sc. in Information Systems from the Technical University of Munich (TUM), where my research focused on Federated Learning. My theses explored various aspects of Federated Learning, including different optimization strategies and deployment on mobile devices. Outside my professional life, I’m an avid football player and an enthusiastic cook.
Core Competencies:
- 🧠 Federated Learning & On-Device ML - Designing and shipping privacy-first training pipelines and edge clients for Apple and Android devices.
- 📱 Mobile ML Engineering - Building Swift and Kotlin clients, integrating on-device ML capabilities, and delivering production-grade mobile experiences.
- 🔬 Applied ML Research - Translating research into production impact through federated optimization work, publications, and thesis-driven experimentation.
- ⚙️ Production Engineering - Implementing robust backend services, APIs, and integration layers across Python, Swift (Vapor), and Java ecosystems.
- 🌐 Product-Focused Web Development - Building modern interfaces with Next.js, React, and TypeScript while preserving usability, performance, and maintainability.
- 🏗️ Architecture & Extensibility - Designing plugin-based systems and integration-ready architectures that support third-party identity providers and enterprise interoperability.
- ✅ Quality & Reliability - Using test-driven frontend workflows, clean abstractions, and pragmatic engineering standards for stable releases.
- 🎨 Design-Aware Engineering - Combining strong UI craft with technical depth to deliver products that are both visually polished and technically sound.
🧰 Overall Tech Stack Summary
The table below summarizes technologies used in my delivered work across Flower Labs, published research, and product-focused engineering projects.
| Pipeline Stage | Tool/Technology | Usage (Project/Course) | Simple Explanation |
|---|---|---|---|
| Federated Learning & Edge ML | Flower Framework (Python) | Flower Labs - development of privacy-first federated learning workflows and framework capabilities | Core platform work for training models collaboratively without centralizing raw data. |
| Flower Swift SDK | Flower Labs + Federated Learning with Swift publication - Apple-device client rollout for federated learning | Enabled iOS and Apple ecosystem devices to participate as first-class federated clients. | |
| Flower Kotlin SDK | Flower Labs - Android federated client implementation | Brought federated training capabilities to Android devices through a production-ready Kotlin client. | |
| gRPC | Flower SDK and federated workflows - model and metric communication between clients and server | Reliable transport layer used for efficient, structured communication in distributed training. | |
| Mobile ML & App Development | Swift / SwiftUI | Xpense iOS project, Flower Swift SDK, and educational app delivery in iPraktikum | Used to build native Apple experiences and mobile-facing ML-enabled products. |
| MLX / CoreML | Apple ecosystem ML integration for federated and on-device workflows | Applied for model execution and evaluation in resource-constrained edge-device contexts. | |
| Kotlin + TensorFlow | Flower Kotlin SDK and Android-side federated learning tasks | Combined native Android development with ML frameworks for production client delivery. | |
| MapKit | Xpense - location-aware features in the iOS app | Integrated native mapping features into the product's finance and usability workflows. | |
| Modeling, Experiments & Research | PyTorch | Flower Labs model experimentation and federated learning prototyping | Primary deep-learning framework for building and validating training experiments. |
| XGBoost | Master thesis - federated tree-based optimization through subsampling techniques | Used to evaluate how tree-based models can be improved under federated constraints. | |
| NumPy | Research code and optimization routines in thesis and project experimentation | Numerical backbone for fast array operations and custom algorithm exploration. | |
| Federated Optimization Methods | B.Sc./M.Sc. theses and publication work around mobile FL performance and sampling strategies | Research translated into practical implementations and documented results. | |
| Backend & Integration Engineering | Python Services | Flower ecosystem tooling and ML workflow integration | Used for orchestration and service-level logic around model training and experimentation. |
| Vapor (Swift Backend) | Xpense - full backend implementation for a production-style iOS companion API | Provided API endpoints and backend infrastructure in a cohesive Swift stack. | |
| Java + Spring + REST | Technolas Perfect Vision - electronic signature system and enterprise middleware delivery | Implemented compliant backend flows and system integration points for business-critical processes. | |
| SQL / JDBC | Technolas Perfect Vision - database triggers and automation logic | Enabled reliable data workflows and process automation in enterprise systems. | |
| Web Product Engineering | Next.js + TypeScript | Portfolio and product web applications, including consumer-facing projects | Delivered modern, typed frontend systems with strong development ergonomics. |
| React + Chakra UI | Public Value Technologies - song voting platform with reusable UI components | Accelerated frontend delivery while maintaining consistency and usability. | |
| Tailwind CSS | Traveldigga and portfolio interfaces | Used to build maintainable design systems with fast iteration cycles. | |
| JavaScript + JSP | Legacy player migration and platform integration at Public Value Technologies | Handled modernization work in mixed legacy/modern frontend environments. | |
| Quality, Architecture & Delivery | React Testing Library + Jest | Public Value Technologies - frontend test coverage for reliability | Improved confidence in releases through component and behavior-level testing. |
| Plugin-based Authentication Architecture | Flower Labs - extensible identity-provider integration in Flower | Designed authentication to support multiple third-party providers without hard coupling. | |
| Middleware Integration | Technolas Perfect Vision - PTC Integrity and CMS interoperability | Connected enterprise tools through stable interfaces and automation bridges. | |
| Git/GitHub Workflow | Cross-project collaboration across ML, mobile, web, and research efforts | Provided structured version control and repeatable collaboration practices. | |
| Publications & Featured Projects | Federated Learning with Swift (Publication) | Co-authored publication on extending Flower for Swift clients and evaluating performance | Formalized engineering and research outcomes in a peer-visible publication. |
| Xpense (iOS + Vapor) | Educational full-stack app for finance management with native mobile UX | Demonstrates end-to-end ownership from client UX to backend services. | |
| Farmily / Traveldigga | Consumer-facing web products emphasizing UI quality and practical user flows | Highlights product-minded engineering and visually polished interface work. | |
| HalmaAI Engine | AI-based strategy project for board-game decision making | Applied algorithmic reasoning and search-oriented thinking in a game domain. |
Work Experience
Machine Learning Engineer @ Flower Labs
May 2023 - present
Developed Flower Swift SDK, the first-ever Federated Learning edge device rollout for Apple devices. Developed Flower Kotlin SDK, a Flower client to perform Federated Learning tasks on Android devices. Designed and implemented user authentication by creating a plugin-based architecture for easy integration of third-party authentication or identity provider into the Flower framework. Conducted various Federated Learning collaborations with other companies under NDA; additional details available upon request.
Tech Stack: Python, PyTorch, Swift, MLX, Kotlin, TensorFlow
Web Developer @ Public Value Technologies
August 2021 - April 2023
Designed and developed a song voting website using ChakraUI and Next.js to create a user-friendly platform with an efficient development workflow. Spearheaded the integration of a proprietary video player into a legacy system, utilizing JSP, Java, and JavaScript to replace the existing player and significantly enhance video functionality and playback performance. Conducted thorough unit testing with React Testing Library and JEST, ensuring the application's reliability.
Tech Stack: Next.js, ChakraUI, React, JavaScript
Working Student @ Technolas Perfect Vision
April 2019 - March 2021
Designed and developed an electronic signature solution using Java with Spring, JavaFX, and RESTful APIs, ensuring an automated and compliant signing experience. Implemented middleware to integrate PTC Integrity with a Content Management System, facilitating efficient data exchange and system interoperability for technical documentation. Developed triggers for a relational database using SQL (Microsoft SQL), JDBC, and JavaScript to automate database-related processes.
Tech Stack: Java, Spring, JavaFX, REST
Projects
Federated Learning with Swift: An Extension of Flower and Performance Evaluation
Co-authored the article that introduces a Swift-based client implementation of the user-friendly Federated Learning framework Flower. This paper is based on my Bachelor thesis titled: "A Prototype Implementation of a Mobile Federated Learning Framework".
- Type: Publication
- Technologies: Python, Swift, CoreML, gRPC
Xpense: User-friendly iOS Application for Tracking and Manage Finances
Xpense is an iOS app designed to help users manage their finances. Used as an educational tool in the iPraktikum program, Xpense serves as a hands-on project for students during a two-week introductory Swift course. It also has a full-fledged Vapor backend.
- Type: Project
- Technologies: Swift, Vapor, SwiftUI, MapKit
Farmily: Connecting Health-Conscious Consumers with Sustainable Farmers
Farmily is a platform that connects health-conscious consumers directly with farmers, making it easy to buy fresh, ethically produced food straight from the source. Farmily enables consumers to support local agriculture and make informed choices about the food they eat.
- Type: Project
- Technologies: React, JavaScript, ChakraUI
Federated Tree-Based Models Optimization through Subsampling Techniques
This research explores subsampling techniques to improve federated gradient boosting models. It compares various subsampling techniques and proposes a new adaptive algorithm capable of selecting the optimal subsampling rate based on data distribution.
- Type: Thesis
- Technologies: Python, XGBoost, NumPy, gRPC
Traveldigga: Your Gateway to Unique and Unforgettable Travel Experiences
Traveldigga is a modern online travel agency designed to inspire and support travelers in exploring unique destinations around the world. Traveldigga offers personalized trip planning, curated hidden gems, and seamless booking for accommodations and tours.
- Type: Project
- Technologies: TypeScript, Next.js, Tailwind
HalmaAI Engine: The Ultimate AI-powered Halma Strategy and Tactics Engine
HalmaAI Engine is designed to play the classic strategy board game Halma with remarkable precision and efficiency, offering users an enjoyable and competitive match. It is capable to analyze the game board, predict optimal moves, and adapt based on their opponent moves.
- Type: Project
- Technologies: React, JavaScript, CSS
Download full CV: Download