This is my short résumé in HTML format. For the full CV PDF click here.
Olaf Lipinski

Olaf Lipinski

Research Fellow in Artificial Intelligence

University of Southampton

About Me

I am a Research Fellow in Artificial Intelligence at the University of Southampton, having recently completed my Ph.D. in emergent communication. I’m currently applying machine learning techniques to maritime engineering challenges, working with industry partners on optimizing car carrier designs through neural networks that estimate wave-induced power requirements. My doctoral research investigated how autonomous agents develop language from scratch. I focused on temporal dynamics in agent communication, designing architectures that enable temporal referencing and improve strategy development. My research examining spatio-temporal relationships in interpretable emergent communication has been published in venues such as NeurIPS and ICLR.

Interests
  • Artificial Intelligence
  • Computational Linguistics
  • Emergent Communication
  • Reinforcement Learning
  • Data-Centric Engineering
Education
  • PhD in Artificial Intelligence, 2020–2025

    University of Southampton

  • BSc in Computer Science, 2016–2020

    University of Liverpool

Skills

Technical
Machine Learning

PhD in Artificial Intelligence.

Python

All publications and side projects coded in Python.

Linux/HPC/SLURM

Worked with SLURM, Linux and HPC both during PhD and at Intel.

VCS/Github

Automation pipeline, CI/CD testing.

PyTorch

Most of my work is in PyTorch (Lightning ⚡️).

Hobbies
Homelabbing

Homelab cluster with Nextcloud, GitLab, IoT, email and others!

Electronic Engineering

Made prototype-level PCBs.

Baking

Recently experimenting with sourdough bakes (focaccia 🍞!).

Experience

 
 
 
 
 
University of Southampton
Research Fellow
February 2020 – Present Southampton, UK
  • Working with industry partners on car carrier design optimization through ML.
  • Implementing neural networks to estimate wave-induced power requirements.
  • Building custom loss functions for maritime experimental data analysis.
  • Initial models accelerate added power predictions for car carrier hulls by 100x vs semi-empirical methods.
  • Expertise: AI/ML, Python, PyTorch, MLOps (MLflow), Scientific ML
 
 
 
 
 
University of Southampton
PhD Student
September 2020 – February 2025 Southampton, UK
  • Thesis: Temporal Dynamics in Emergent Communication
  • Supervised by Prof. Timothy Norman, Prof. Adam Sobey, and Prof. Federico Cerutti.
  • Investigated the effect of time on communication protocols, improving agents’ strategy development, and convergence speed by up to 50%.
  • Designed an architecture necessary for the emergence of temporal referencing in agent communication, enabling efficient communication.
  • Developed human-interpretability measures for an emergent language containing local spatio-temporal relationships.
  • Built end-to-end ML pipelines for agent training and evaluation.
  • PhD defended with no corrections.
  • Expertise: AI/ML, Python, PyTorch (Lightning), Ray (RLLib).
  • Professional Skills: Research impact, research communication, academic collaboration.
 
 
 
 
 
Intel Corporation
HPC Technical Engineer Intern
January 2018 – January 2019 Swindon, UK
  • Managed client-facing HPC and Cloud facilities, ensuring optimal performance.
  • Collaborated on pre-production hardware projects, developing technical solutions.
  • Created deployment documentation and automation scripts to improve system efficiency.
  • Expertise: HPC, Cloud, Bash, Linux, Server Hardware.
 
 
 
 
 
VLDB Solutions
Technical Consultant
June 2016 – April 2028 Liverpool, UK
  • Developed Bash and SQL scripts for customer data analysis using AWS services.
  • Progressed from intern to part-time role across multiple projects.
  • Expertise: Big Data, Amazon AWS (EC2, S3, Redshift), SQL, Bash.

Publications

(2024). Speaking Your Language: Spatial Relationships in Interpretable Emergent Communication. NeurIPS.

PDF Cite Code

(2023). On Temporal References in Emergent Communication. arXiv.

PDF Cite Code DOI

(2022). Emergent Password Signalling in the Game of Werewolf. In EmeCom Workshop at ICLR 2022.

PDF Cite Code