I’m a senior research engineer specializing in large language models. At Luminance, I develop, train and deploy LLMs for various legal tasks and scenarios. I occasionally participate in research projects at The Context Lab.
I hold a first-class honours BSc in Computing Science from the University of Aberdeen. My research on context effects in dialogue models has been featured in conferences like CoNLL and EMNLP.
My software engineering experience spans healthcare, defence, drone management, telephony, and safety consulting. I’ve built systems for business management, drone shows, task automation, telephony servers, area surveying and many more.
My contributions to AI and engineering have been recognized in leading journals, conferences and competitions.
| Links |
|---|
| 🔗 LinkedIn/molnaron |
| ⚙️ GitHub/Arotte |
| 🔬 Google Scholar |
| 🐔 X/aronmolnr |
| 🤗 HuggingFace/Arotte |
Experience
AI/LLMs
- Senior Research Engineer @ Luminance AI – Developing custom LLMs for various legal tasks and scenarios.
- Research Assistant @ University of Aberdeen – Investigating and improving dialogue agents.
- AI Developer @ HyperDAP – Made algorithms to automatically detect and forecast potential oil well failures based on temperature sensor readings.
Software Engineering
I architected and designed a suite of software systems for the University of Aberdeen, IT-Pro, HyperDAP, CollMot and Ericsson. These involved both client-facing and internal and front- and back-end systems. A few examples:
- Business and inventory management systems for the NHS and defence industry suppliers.
- Drone fleet management software for drone shows.
- Internal task automation tools.
- Telephony application servers responsible for managing millions of daily calls.
- Area surveying mobile applications for safety consultancies.
Education
BSc Computing Science (2019-2023)
- University of Aberdeen
- Graduated with 1st class honours.
- Thesis: Interpreting Transformer Attention During Next Utterance Dialogue Generation from Human Contexts.
BSc Electrical Engineering (2017-2019)
- Budapest University of Technology and Economics
- Built projects like a sensor that can detect the tide of the moon, or a concept seismic shield against earthquakes.
Research
At The Context Lab (University of Aberdeen), I investigate LLM behaviour in dialogue settings.
Publications:
Attribution & Alignment: Investigating the Effects of Context on Model Utterance Production and Comprehension in Dialogue, Molnar, A., Jumelet, J., Giulianelli, M. & Sinclair, A. Appeared in CoNLL 2023, presented at EMNLP in Singapore. Arxiv, GitHub.
Analysis of a ferrofluid core differential transformer tilt measurement sensor, T Medvegy, A Molnar, G Molnar, Z Gugolya. Journal of Magnetism and Magnetic Materials 428, 189-193. Project Page.