Scientific software development
I am currently a scientific software developer at King’s College London within a newly established central research software engineering team. Generally, I aim to write science-enabling software that is more reproducible, shareable, and scalable, and allows to more easily build upon existing research. For me, being a scientific software developer means to contribute to a larger project – that of improving science. Currently, I am involved in co-building a gen-AI-based, to-be-commercialised app to enable the collection of rich but narrowly scoped qualitative data at scale. This work operates across human-computer interaction research and product development where we build production-ready software while requirements are still being discovered through research - a challenge distinctive to scientific software. You can read about a pilot study using a prototype of the app in this paper.
Research in computational neuroscience, complex systems and technical AI safety
I have recently completed a PhD in computational neuroscience and complex systems at the Sussex Centre for Consciousness Science, University of Sussex. My doctoral research (comprising three to-be-published studies - watch out for preprints!) addresses both conceptual/philosophical and empirical questions about multi-scale relationships in complex systems, focusing on formal, information-theoretic approaches to emergence and complexity. Conceptually, I co-developed a framework that rethinks emergence as a non-binary, multi-dimensional construct. Empirically, I systematically validated a broad range of emergence and complexity measures across multiple simulated time-series models and macro-scales to inform empirical research on time-evolving systems. I also examined multi-scale measures in learning systems employing Approximate Bayesian Inference, a method from machine learning. Beyond this, I’ve leveraged my expertise in complexity science/emergence research for AI safety within a PIBBSS fellowship. My aim was to bring together various strands of research (philosophical, formal/mathematical and empirical) on the concept of emergence to inform and bring progress on research in AI capabilities. More specifically, I intended to explore whether gained insights can be leveraged for evals-type of work to produce a deployable “emergence-assessment pipeline” for assessing AIs w. r. t. their emergent capabilities. Here’s a talk on my AI safety research project.
AI/ML and datascience experience
My work with AI/machine learning spans several domains beyond the technical AI alignment and LLM-based software development work mentioned above. During my PhD, I applied unsupervised ML techniques to derive candidate emergent macro-scales in simulated time-series models. Prior to my doctoral research, I used unsupervised machine learning to investigate robotic hand control via motor imagery and experiences of embodiment in an EEG-based brain-computer interface study. Following this, I investigated the dynamics of integrated information (a complexity measure) in black-box variational inference, a foundational technique used in ML for approximating intractable posterior distributions in Bayesian inference. This study formed the basis for investigating multi-scale measures in learning systems employing Approximate Bayesian Inference as part of my PhD.
Beyond machine learning applications and foundations, my data science work extends to statistical analysis and research methodology across diverse domains. As a research data scientist in medical education research, I analysed questionnaire data to inform the development of practice- versus competence-based assessment procedures in medical education. This work resulted in a co-authored publication applying statistical methods to inform educational policy questions. Through my (doctoral) research, I’ve developed deep expertise in information theory (an advanced branch of statistics), time-series modelling, and Bayesian methods. Early in my career, I investigated the generalizability of quantitative studies in experimental (neuro)psychology through analysis of deployed sampling methods – work that highlighted critical issues in external validity.
Open science research software advocacy
As a 2023/2024 Software Sustainability Institute (SSI) Fellow, I sent a strong signal for the institute’s mission of “better software, better research” by delivering research software development workshops to research communities I actively participate in. You can check out (and re-use) material for a research software enginnering tutorial in Python that I’ve created for the Artificial Life conference 2023 in the context of my fellowship. As part of my PhD project, I developed ComplexPy – a Python library (still work in progress) designed to make different complexity and emergence measures easily applicable and comparable across datasets in an intuitive, comprehensive way. This development work was supported through my participation in the Open Seeds mentoring and training program, where I also served as a mentor to other open science and open software enthusiasts. In the context of my EEG-based BCI study, I’ve developed a bootstrapping function for two-way mixed effects ANOVA in R. I’ve been an active contributor to the open science ecosystem through regular participation in Brainhacks, some of which have resulted in open-source software contributions. All code from my past and ongoing research is published and documented.
Education and mentoring
I’ve been engaged in knowledge sharing across diverse educational contexts and audiences. In my current role as a scientific software developer at King’s, I have provided support for HPC training. During my academic research, I’ve taught courses spanning ML fundamentals, Algorithmic Data Science, Practical Philosophy, and EEG methods, and supervised machine learning theses. I’ve delivered research software engineering tutorials and served as a mentor multiple times in the Open Seeds mentoring and training program, helping participants to open-source their software projects. I’ve also taught German to refugees, and was, early in my career, involved in an “Idea-Of-Man” seminar that approached psychological topics from philosophical, political, and sociological perspectives, coaching students to critically reflect on these interdisciplinary connections.
Interdisciplinarity and seeing the big picture
Most of my research has been hugely interdisciplinary, intersecting mathematics, machine learning, neuroscience, as well as philosophy. My doctoral work in particular required extensive bridge-building between abstract mathematical measures and their conceptual meanings.
I have been the main organiser of a symposium titled Rethinking Computational Approaches to the Mind - Fundamental Challenges & Future Perspectives, which brought together researchers to examine critical questions about computational approaches in the mind sciences. Have a look at the recorded talks and panel discussion which included, amongst others, Melanie Mitchell, Konrad Kording, and Gaël Varoquaux.
I thought much about metascientific questions, one outcome of which is an extensive article-like post on collaboration in science, arguing that shifting from individualistic to collaborative work in academia can improve scientific progress and increase well-being. This is particularly crucial in the mind sciences where research methodology is especially challenging.
Currently, I’m exploring the fundamental tension between Buddhist non-attachment and effective action in a written piece – how Buddhism teaches us to loosen our grip on views to reduce suffering, while meaningful action requires (some minimum) clinging to views. I examine the cycle of loosening perspectives during assessment, then solidifying around synthesized views to generate energy for action, and address Buddhism’s limitations in providing concrete guidance for what we should actually do beyond responding “skillfully.”
Psychology and human rights work
I have a background in psychology and worked as a psychologist for unaccompanied refugee minors at Bahia. I also engaged in pedagogical psychology work at Children for Tomorrow, focusing on supporting vulnerable young people. During an internship at Equality California in Los Angeles, I contributed to public education projects including The Breakthrough Conversation and Health Happens With Equality, while also initiating a project addressing biphobia.
Volunteering
I worked as an environmental sustainability volunteer with GrowNYC’s Zero Waste and Greenmarket programs in New York City, supporting local food systems and waste reduction efforts. Earlier, I was active with Amnesty International in Hamburg, Germany, until 2012, contributing to human rights advocacy work.
Art and the outdoors
I engage in photography and share my work on a photo blog (discontinued - new work will be posted elsewhere) as well as on Instagram. I also post about (ultra mountain and sky)running, the outdoors more generally & music/culture in the UK/around the world in another account. In this blog/personal website, I share writings on various topics.