Profile
Balázs László
I work across AI safety research, governance, and community building. My research explores how language models represent linguistic structure and how those representations shape behaviour.
I’m Hungarian, currently based at CEEALAR in Blackpool, UK. I’m interested in connecting technical understanding with practical work that helps people and other animals.
The longer commitments, the shorter experiments, and the learning along the way.
Preview CV
Work & ongoing commitments
Oct 2025–present
AI Researcher
Algoverse AI Research
Researching steganography detection through mechanistic interpretability, with a focus on experimental design and model analysis.
Aug–Nov 2026
Mentor
Sentient Futures Incubator
Supporting fellows in shaping research projects, including work on synthetic data, linguistic differences, and animal-aligned value generalisation.
Apr–Sept 2026
AI Integrations
LINAK · Operations Excellence & R&D
Developed technical and operational plans for integrating AI into company workflows, and led workshops and presentations.
Mar–May 2026
Research Fellow
Sentient Futures Incubator
Designed a welfare-first framework for monitoring and mitigating zoonotic disease risk, and began developing Disease Free Futures.
Sept 2024–present
Leader
Next Generation Hungary · Erasmus+
Contributed to international youth initiatives on climate change and youth empowerment, including projects in Portugal, Greece, and Lithuania.
Jul 2024–Jul 2025
Secretary General
JEF Hungary
Directed international youth projects, conferences, and policy discussions; secured and administered grants; represented Hungary in the Federal Committee.
Sept 2022–Jan 2023
Programming Educator
Logiscool · Tatabánya
Taught Python to secondary-school students through interactive exercises and project-based lessons.
Selected projects & sprints
May–Jun 2026
Linguistic representations in language models
BlueDot Impact · Technical AIS project sprint
Investigated whether low-level linguistic features are represented, sparsely separable, and suitable targets for causal interventions. Established the experimental foundation for follow-up research.
Sept 2025
CBRN benchmark evaluation
Apart Research · Research sprint
Developed Python prototypes to investigate weaknesses in CBRN AI benchmarks. Our team placed second globally.
Dec 2024–Mar 2025
AI governance & interdisciplinary research
TUDOK
Developed a proposal for reducing language-model bias through structured databases informed by the EU AI Act. Received international second-place awards for the AI governance proposal and an MLM participation study.
Jun–Nov 2024
Student participation in MLM networks
BGE TDK & OTDK
Co-authored a mixed-methods study combining quantitative analysis with qualitative research. Advanced to the national round and received a special award.
Jun–Sept 2024
Competition robotics
FIRST Global Challenge · Athens
Co-built and programmed a robot using Python and sensor-based control. Our team ranked 34th among 193 international teams.
Education
2025–2026
Mechatronics studies
University of Southern Denmark
Studied mechatronics; founded and led SIRAIS (Student Initiative for AI Safety Research), and worked on inventory automation for the ARC Rocketry Club.
Jan 2026
Computer Science admission offer
University of Cambridge
Offered admission to the Computer Science Tripos; unable to take up the place due to insufficient funding.
Training
Sept 2026
Technical ML4Good Bootcamp
Greens Norton · UK
Intensive training in machine learning, AI safety, and applied research methods through technical exercises and collaborative projects.
Aug–Oct 2025
AGI Strategy
BlueDot Impact
Studied governance, safety, and long-term planning for advanced AI systems.
Jan 2022–Jun 2025
Talent programme
Mathias Corvinus Collegium
Multidisciplinary training across STEM and social sciences, with seminars and conferences involving academics, entrepreneurs, and policymakers.
Aug–Sept 2023
AI Safety Technical Fellowship
AI Safety Hungary
Explored AI alignment, interpretability, machine learning, and research methodologies.
Tools & languages
Technical Python, R, C, C++, SQL, LaTeX · Git, MATLAB, NX CAD, Power BI
Languages Hungarian · English · German · Danish
Behind the entries
A little more about me
Research & questions
I’m particularly interested in the relationship between language and model behaviour: how relatively small linguistic differences are represented inside a model, which mechanisms they share, and whether we can intervene on one property without changing others.
My work has included research on linguistic representations, benchmark evaluation, and mechanistic interpretability. I also mentor research projects through Sentient Futures, including work at the intersection of language models and animal welfare.
Into the research compartment ↗Building communities
Alongside research, I’ve spent time organising groups, teaching, and bringing people together. At the University of Southern Denmark, I founded SIRAIS, the Student Initiative for AI Safety Research.
As Secretary General of JEF Hungary, I worked on international youth projects, grants, conferences, and policy discussions. I’ve also taught Python and contributed to Erasmus+ initiatives through Next Generation Hungary.
Beyond work
I write poetry and enjoy calisthenics. Both give me something quite different from research: room to experiment, pay attention, and work patiently at something that matters to me.
This website is a home for more than a professional biography. The other compartments hold some of the writing, interests, and ideas that don’t fit neatly into a CV.
Say hello.
For research, collaborations, or a conversation.