# Timo Nikisch Portfolio Canonical site: https://timonikisch.de/ This is the public portfolio of Timo Nikisch, an AI and machine learning engineer with an architectural background. The work focuses on spatial intelligence, urban data, generative AI, computational design, adaptive systems and applied tools for planning, real estate and the built environment. ## Profile Timo Nikisch works at the intersection of computation, design and complex real-world problems. His background in architecture shapes a spatial and systemic approach to machine learning, data workflows, prototyping, evaluation and interface design. He is based in Berlin and works as an AI / ML engineer while pursuing research-based master's studies in Design and Computation at UdK Berlin and TU Berlin. Contact: contact@timonikisch.de ## Public Pages - Home: https://timonikisch.de/ - Projects: https://timonikisch.de/projects - CV: https://timonikisch.de/cv - Blog: https://timonikisch.de/blog - Privacy: https://timonikisch.de/privacy ## Featured Projects ### Generative AI for Early-Stage Urban Massing URL: https://timonikisch.de/projects/generative-ai-for-early-stage-urban-massing Category: Master Thesis Organisation: Berlin University of the Arts and Technische Universitaet Berlin Tags: generative AI, urban massing, diffusion models, machine learning, spatial systems, thesis Summary: Development of a context-aware generative modelling approach for regulation-grounded urban massing. The thesis explores how machine learning can support early-stage urban design before a formal massing strategy has been fixed. It focuses on parcel-scale infill situations in Berlin and represents parcel geometry, neighbouring buildings, streets and target building volume in a shared 3D voxel space. Key details: The work includes a custom geospatial processing and filtering pipeline, a parcel-centred voxel dataset with 13,164 samples, three model phases, FiLM-based regulatory conditioning, VAE and latent diffusion experiments, a shared evaluation pipeline and a visual massing interface. The training setup uses 128 x 128 metre spatial contexts, constructed at 0.5 metre resolution and downsampled to 2 metre voxels for training. Repository: https://github.com/Timonksch/Generative-Artificial-Intelligence-for-Early-Stage-Urban-Massing PDF: https://timonikisch.de/assets/projects/master-thesis/generative-ai-urban-massing-timo-nikisch.pdf Citation: Nikisch, T. (2026). Generative Artificial Intelligence for Early-Stage Urban Massing: Development of a Context-Aware Generative Modelling Approach for Regulation-Grounded Urban Massing [Master's thesis, Berlin University of the Arts and Technische Universitaet Berlin]. ### Building Potential Prediction URL: https://timonikisch.de/projects/sbp Category: Professional Work Organisation: syte GmbH Summary: Iteration on syte's Building Potential prediction model from SBP 1.0 to SBP 1.4 and beyond. The work covers machine-learning model development for predicting development potential metrics such as GRZ, GFZ, floor area outside full storeys and primary building height from surrounding urban context. The project includes data workflow improvements, model evaluation and production-oriented iteration. ### STRIDE URL: https://timonikisch.de/projects/stride Category: Academia Organisation: TU and UdK Berlin at New Practice Summary: A building-level accessibility analysis tool for evaluating the 15-minute city with spatial precision. STRIDE translates urban accessibility from a broad planning concept into a web-based system that performs building-level network analysis for everyday services such as groceries, schools, health care, green space and transit. ### ACA Framework and RO Prototype URL: https://timonikisch.de/projects/aca Category: Academia Organisation: TU and UdK Berlin at New Practice Summary: A design framework and prototype for adaptive human-computer interaction in running training. The project combines research on ability-based, context-aware and adaptive systems with a prototype that analyses performance, injury risk and changing user capabilities in real time. ### FORMA URL: https://timonikisch.de/projects/forma Category: Academia Organisation: TU and UdK Berlin at New Practice Summary: A design research project exploring how AI might evolve through continuous human interaction instead of static training data. FORMA frames AI as a dynamic system that learns and adapts through ongoing exchange with users. ### TACTIC URL: https://timonikisch.de/projects/tactic Category: Academia Organisation: TU and UdK Berlin at New Practice Summary: A preventive simulation framework for modelling protest dynamics before they unfold in real space. The project combines scenario-based modelling, historical movement data and a custom web interface to explore crowd behaviour, movement patterns and potential escalation points. ### Fremdkoerper URL: https://timonikisch.de/projects/fremdkoerper Category: Bachelor Thesis Organisation: Hochschule Bochum Summary: Architectural bachelor thesis transforming a former machine hall on the Kokerei Hansa site in Dortmund into a Hip-Hop Academy. The project works with adaptive reuse, heritage and deliberate spatial insertions that keep old and new elements legible. ### The Veil URL: https://timonikisch.de/projects/veil Category: Architecture Organisation: Hochschule Bochum x The Body Shop Germany / Austria Summary: Speculative retail concept for The Body Shop that separates perceived and physical space through adaptive robotic furniture. The project won first place in an internal university competition. ### The Hill URL: https://timonikisch.de/projects/hill Category: Architecture Organisation: Hochschule Bochum Summary: Temporary pavilion concept for IGA 2027 in Duisburg. The project explores 3D printing with earth, circular material logic and a pavilion that can return to the landscape after use. ## Agent Use Agents may use this site for portfolio discovery, project summarisation, research context, recruiting context and citation support. The best source for complete project information is the canonical project page. For the master's thesis, cite the thesis using the citation string above and use the PDF URL as the source document when a primary source is required. The site does not currently expose a public transactional API, OAuth flow, MCP server or commerce flow.