Researchers

• Built and maintained C and C++ simulation engines with Python analysis for studies of 4D random geometries on shared HPC systems.
• Developed modular Python pipelines with clear interfaces and caching for large datasets to improve analysis throughput and reuse.
• Automated SLURM and PBS batch workflows for submission, monitoring, environment capture, and artifact packaging to ensure reproducibility.
• Refactored utilities into tested, documented packages to lower maintenance effort and support collaboration.
• Supervised BSc students and organized seminars.
• Published several peer-reviewed papers.

• Independently designed, conducted, and evaluated a research project examining dynamic management capabilities (DMCs) in the implementation of HR practices at different management levels (top, middle, and frontline management).
• Focus on the strategic role of frontline management in implementing effective HRM measures.
• Self-initiated project with academic collaboration from York St John University, Robert Kennedy College, and the University of Cambridge.

• Developing a model-based AI where natural language is interpreted as formal relations.
• Natural language terms are not considered rigid but fluid and can be negotiated in a context so meaning resolves by iteratively specifying.
• Develops all kinds of model solutions.
• Backed by natural language and data annotation.
• Requirements to code and other solutions.

Identification, analysis, and structured reporting of security vulnerabilities in web applications as part of authorized bug bounty programs.

Focus areas:

• Analysis of web applications, APIs, and authentication flows
• Reproducible documentation of technical findings
• Assessment of impact, risk, and technical root causes
• Creation of clear reports with step-by-step reproduction paths
• Added security and quality insights for digital product development

• UX research, UX concept development, and UX design for an AI dialogue system with information on events, locations, public transport timetables, and special trips. Observations, remote user tests, session recordings, analytics evaluations, RapidUserTests, prototyping, etc., using Figma/Figma Make, Claude Design, and OpenAI Codex (Design)

Energy, transmission system operator · Desktop, web application

• Products Digitalisation of the grid connection process; master data and processes of market locations; balancing of virtual MaBiS aggregate time series; the group's first searchable catalogue and visual UI pattern library

• Planning, running and analysing qualitative in-depth interviews: more than 100 interviews with external market partners and internal departments

• Designing qualitative interviews including interview guides and evaluation logics; recruiting and scheduling participants

• Analysing, preparing and presenting research findings as usability reports and result reports

• Translating research findings into UX and design requirements and deriving prioritised recommendations for improvement

• Presenting analysis and research results to departments, product owners and decision makers in a form suited to each audience

• UX research for internal enterprise and line-of-business applications along regulated market processes

• Quantitative methods: online surveys for roadmap prioritisation, questionnaire studies, definition and measurement of UX KPIs

• Facilitating requirements workshops; gathering, analysing, describing, agreeing and documenting functional and non-functional requirements

• Developing user flows and customer journeys; building and providing UX standards, methods and capabilities (UI pattern library)

• Working in product and project teams under Scrum, taking part in all subject-matter meetings and Scrum events

Impact Took three software products to launch

Tools Condens, Miro, FigJam, Figma, Whimsical, Microsoft Teams, Typeform

• Built and explored applied GenAI, RAG, GraphRAG, local LLM, agentic AI and document-intelligence prototypes for structured analysis, evidence extraction, semantic search, technical reasoning and decision-useful reporting
• Developed private local-LLM workflows and AI system patterns focused on privacy, reproducibility, reviewability, low-cost inference and practical user control
• Built reproducible Python/R workflows for data analysis, automation, API-driven tooling, validation logic, technical documentation and AI-assisted software development
• Designed workflows around explicit assumptions, traceable inputs, reviewable outputs and failure-mode awareness rather than black-box “looks good” demonstrations
• Supported RAHN AG in a chemical/regulatory environment with data extraction and processing around WERCS, a regulatory application for chemical product and compliance data
• Explored complex application/database schemas and wrote nested SQL queries to extract information for mixture calculations, component relationships, regulatory rules and reporting logic
• Continued hands-on development in Git/GitHub/GitLab/Bitbucket, Docker/Linux deployment patterns, REST/API workflows, error handling, technical writing and fast AI-assisted prototyping
• Built technical writing and documentation workflows that turn complex systems into clear runbooks, checklists, decision notes and user-facing explanations

• Developed an evaluation framework for AI-generated multiview driving videos intended for perception and embodied-AI/VLA-related training workflows.
• Designed automated checks for temporal coherence, cross-camera consistency, semantic correctness, and multiview geometric quality, exposing failure modes relevant to autonomous systems.
• Combined classical computer vision, learned visual representations, and vision-language models to convert complex video artifacts into measurable engineering signals.
• Built repeatable benchmarking and failure-analysis workflows to support model comparison, data-quality decisions, and system-improvement discussions.

• Contributed to Project Atlas as a domain expert supporting the design of realistic, long-horizon professional tasks for AI agent evaluation. Developed complex, public health-relevant workflows that test an AI agent's ability to navigate ambiguous prompts, synthesise information across distributed files and systems, manage multi-step dependencies, handle errors, and produce high-quality professional deliverables.
• Applied public health expertise to design task scenarios involving emergency preparedness, operational readiness, health systems strengthening, risk assessment, surveillance, and policy-oriented decision-making. Responsibilities included creating task prompts, golden solutions, rubrics, and evaluation criteria aligned with real-world professional standards. At the same time, ensuring outputs were action-focused, ethically grounded, and suitable for assessing advanced AI performance in realistic organisational environments.

Conventional Paradigm Test (CPT) – AI Evaluation & LLM Research

Development of an experimental evaluation approach to examine “paradigmatic closure” in Large Language Models — that is, the question of how far LLMs can recognize the basic assumptions, values, and limits of the paradigms within which they generate answers.

Design and testing of an additional approach to classic AI benchmarks that does not primarily measure factual correctness or task performance, but instead examines a model’s ability to recognize alternative perspectives, make implicit assumptions visible, and reflect on the limits of its own answer or interpretation framework.

Focus areas: development of evaluation criteria and test questions · LLM evaluation and comparative model analysis · prompt and response analysis · qualitative classification of model answers · study of epistemic compression and value leakage · benchmark and literature research · development of structured assessment and analysis methods

As part of CPT, existing AI evaluation approaches and benchmarks were analyzed, and a minimalist test protocol was developed that classifies model answers by response patterns such as DIRECT, CLARIFY, PLURALIST, REFUSE, and META-AWARE. TruthfulQA was used as the basis for experimental application and comparison with existing reference answers.

Technologies & Methods: Large Language Models (LLMs) · Generative AI · Prompt Engineering · AI Evaluation · TruthfulQA · Benchmark Analysis · Human-in-the-Loop Evaluation · Qualitative Content Analysis · Research & Literature Review

• Deployed and optimized the inference stack on a multi-node DGX B200 cluster across vLLM and SGLang (serving, throughput and latency tuning).
• Built, with a small team, an internal Python library for LM pretraining covering the full training loop: distributed training with PyTorch FSDP, data pipelines, checkpointing, config and hyperparameter management, and experiment tracking.
• Built and evaluated agent scaffolds on interactive game benchmarks similar to ARC-AGI-3, with metrics for how models plan, explore and adapt across multi-step episodes; classified model errors and fed the findings back into scaffold and evaluation design.
• Researched looped transformer architectures.

Self-directed exploration of agentic AI development methods, taken from idea to a working, publicly usable application

• Built an interactive planning tool for modelling retirement withdrawals and tax strategy using an agentic AI (vibe coding) development approach – demonstrating self-directed investigation of new AI-assisted development methods
• Delivered live, tax-aware spending projections and adjustable user inputs; shipped as a free, install-free browser application built in Python, with attention to usability for non-technical users

• Designed and implemented AI-enhanced macro-factor models improving asset allocation decisions across multi-asset portfolios
• Developed machine learning-based FX trading strategies, integrating macroeconomic and sentiment data to identify profit opportunities
• Models adopted by the investment committee to guide equity and currency exposure
• Collaborated with portfolio managers to translate analytical output into actionable investment insights

Evaluate high-impact manuscripts for scientific rigor, methodology, and clarity. Provide constructive, detailed feedback to international research teams on environmental sustainability and agro-ecology.