cv

Machine Learning Researcher

Centre for Credible AI

• Investigating whether multi-agent LLM systems deliver genuine capability gains or primarily act as additional test-time compute, using mathematical reasoning tasks as the main evaluation setting.
• Teaching a student research group and mentoring its mini-project on emergent misalignment, which builds on Model Organisms for Emergent Misalignment (Turner et al., 2025).
• Member of a red-teaming group on AI control and LLM-based agents; exploring inoculation prompting vs. steering-vector methods and rule-breaking tendencies in LLM agents.

Machine Learning Researcher

University of Warsaw

• RL for Scalable Oversight via LLM Debate. Extended Training Language Models to Win Debates with Self-Play Improves Judge Accuracy to study RL training in multi-agent debate as a scalable-oversight method.

Machine Learning Engineer

MIM Solutions

• Contrastive Learning for Linking Tweets to Events (ERC-funded). Led a research initiative supervised by Prof. Sankowski developing BERT-based dual encoders to link Twitter posts with news articles; owned experimental design, modelling, and analysis. Preprint: https://arxiv.org/abs/2312.07599.
• Automated Counter-Speech to Mitigate Hate Speech (EMNLP Findings 2024). Took over a stalled student project as lead after two rejections: introduced a novel engagement metric, added ethical analysis, and rewrote the manuscript, leading to acceptance. Preprint: https://arxiv.org/abs/2311.16905.
• Research Infrastructure. Trained, evaluated, and fine-tuned deep learning models on GPU clusters using Hydra for configuration, Optuna for hyperparameter optimization, and Neptune for experiment tracking.

Scout Leader, Board Member - Mazovian Region

Polish Scouting Association (ZHR)

• Board Member overseeing quality assurance of camps for 2,000+ youth annually; directly managed camps of 120 participants with a 25,000 EUR budget, and mentored 30 leaders.

PhD Student

Centre for Credible AI, Warsaw University of Technology, Poland

Research focus: multi-agent safety

M.Sc.

University of Warsaw, Poland

Machine Learning

• GPA: 4.58 / 5
• Thesis: Linking news to tweet cascades with a contrastive learning approach

B.Sc.

Warsaw University of Technology, Poland

Computer Science

Interpretable Semantic Gradients in SSD: A PCA Sweep Approach and a Case Study on AI Discourse

ACL 2026 Findings

Authors: Plisiecki, Leniarska, Piotrowski, Zajenkowski.

LLM-generated Responses to Mitigate the Impact of Hate Speech

EMNLP Findings 2024

Authors: Podolak, Lukasik, Balawender, Ossowski, Piotrowski, Bakowicz, Sankowski.