Systems thinking
I care about how components, data and developer workflows fit together — not just isolated functions.
I'm Sebastian Mischke, a backend-focused software engineer working across AI-powered document processing, developer tooling, automation and industrial software. I like systems that are understandable, reliable and pleasant to work with.
About
My work has moved through deep-learning research, full-stack industrial applications, automotive toolchains and, today, AI-driven document processing.
That range made me comfortable moving between layers: understanding the domain, designing an API, building an automation pipeline, debugging a build system, shaping data models and occasionally jumping into the frontend when that's what gets the product over the line.
I care about how components, data and developer workflows fit together — not just isolated functions.
Build pipelines, code generation, ingestion flows and tooling are recurring themes in my work.
Readable solutions, useful abstractions and technology chosen for the problem rather than the trend.
From computer vision to simulation, graph data and UI work — I enjoy learning enough to build things properly.
Experience
elevait GmbH & Co. KG · Dresden
Backend-focused development for AI-powered document processing: Python services and FastAPI APIs, automated ingestion and processing flows with Apache NiFi, document extraction models, data storage and traceable relationships across MongoDB and graph data. Also contributing to React functionality and operational reporting.
FDTech GmbH · Chemnitz
Framework and toolchain engineering in automotive development. Worked with Python and C++ on build systems, automation, release and deployment pipelines, testing and a large monorepository. Projects included architecture parsing and code generation, simulation tooling and datatype specification.
HTW Dresden
Researched and evaluated deep-learning approaches for emotion detection in images and video. Built preprocessing and augmentation pipelines, benchmarked CNN architectures and explored temporal filtering for more stable video predictions.
Saxony Media Solutions GmbH · Dresden
Full-stack development for advanced planning and scheduling software across semiconductor, aviation and production-planning use cases. Built Java/Hibernate and Python backends, REST APIs, database integrations and Angular/TypeScript interfaces.
Bachelor-thesis internship: deep-learning feature recognition in satellite imagery using Python and Keras.
Tutor for first-semester Media Informatics students, focused on mathematics and programming foundations.
Student research assistant applying neural networks to semantic classification of occupancy-grid maps.
Selected work
Deep learning · Computer vision
A complete pipeline for recognizing buildings and properties in satellite imagery: automated image acquisition, training-data tooling, CNN experimentation and result visualization.
European R&D · Industry 4.0
Work on process virtualization and agent-based simulation for industrial planning and semiconductor supply chains within a large European research initiative.
Mobile · Architecture
Android event application built with Xamarin and C#, with MVC separation, local data and image management plus swappable external synchronization connectors.
Open source · Python
Upstream contributions including refactoring, version-aware updates and support for additional addon providers in a lightweight Python utility.
Toolbox
Languages
Backend & data
Tooling & automation
Frontend & ML
Education
Hochschule für Technik und Wirtschaft Dresden
Hochschule für Technik und Wirtschaft Dresden