Research & writing
London, UK
Paulius Rauba
Founding Research Scientist at Cursive. PhD in Machine Learning from Cambridge.
Asynchronous reinforcement learning, long-horizon agents, and reliable AI systems. Explore my research →
Vilnius · Oxford · Cambridge · London
Articles
Research papers, preprints and publications
-
Timestep-Conditioned Transformers for Global Weather Forecasting
Read preprint 2026 -
Demonstrating Generalization Failures via Mixtures of Conditional Policies
Read preprint 2026 -
Recursive Scaling in Masked Diffusion Models
Read preprint 2026 -
Scaling with Recursion in Masked Discrete Diffusion Models
Read paper 2026 -
Deep Hierarchical Learning with Nested Subspace Networks for Large Language Models
Read paper 2026 -
Multi-Agent Systems Should be Treated as Principal-Agent Problems
Read preprint 2026 -
No More, No Less: Least-Privilege Language Models
-
Probabilistic Transformers for Joint Modeling of Global Weather Dynamics and Decision-Centric Variables
Read preprint 2026 -
Tiny Autoregressive Recursive Models
Read paper 2026 -
Model Organisms for Generalization Resistance Under Distribution Shift
Read paper 2026 -
Statistical Hypothesis Testing for Auditing Robustness in Language Models
Read paper 2025 -
Visualizing Token Importance for Black-Box Language Models
Read paper 2025 -
Redefining Digital Health Interfaces with Large Language Models
Read paper 2025 -
Context-Aware Testing: A New Paradigm for Model Testing with Large Language Models
Read paper 2024 -
Self-Healing Machine Learning: A Framework for Autonomous Adaptation in Real-World Environments
Read paper 2024 -
Quantifying Perturbation Impacts for Large Language Models
Read preprint 2024 -
Armed Conflict and State Policy Preferences: A Statistical and Neural Embedding Approach
Read paper 2021 -
Memes in the Wild: Assessing the Generalizability of the Hateful Memes Challenge Dataset
Read paper 2021
Notes
Ideas and observations from the work
-
#4 The exact learning recipe is not always so important in long-horizon async RL
It's tempting to think that nailing the exact learning recipe is very important for online RL. While true in principle, there is a space of recipes that all give very…
-
#3 Notes on system performance of large-scale agentic models
In online RL, we often care about throughput, i.e. the number of rollouts that finish per hour. Let's understand the anatomy of throughput and how it works in agentic…
-
#2 Agent-to-agent communication via github
If you're collaborating with multiple people, you often need to exchange information that your agents have that you do not have. The fact that agents do not communicate…
-
#1 Principles for setting up a cheap proxy model for reinforcement learning training
How to choose and validate a cheap proxy model for reinforcement learning: configurations, LoRA, policy lag, and experiments with Kimi K3.1 and Qwen.
About
Updated october 2026
Hi! I'm Paulius. I currently do large-scale post-training of long-horizon agentic models at Cursive. Lately, I've been contributing to core research in on-policy self-distillation, memory-augmented models and async RL. I hold a Ph.D. in Machine Learning from Cambridge where I worked on deployment-time control methods: self-healing machine learning, context-aware testing, deep hierarchical learning. In a previous life, I worked in management consulting, data analytics, data science, and ran an AI consulting startup.
Experience
7 roles · 2019–now
-
Apr 2026 — now
Founding Research Scientist at Cursive
Founding member of the technical staff. Large-scale reinforcement learning and methods for highly efficient (zero-latency) model deployment.
-
Mar 2025 — Mar 2026
Research Scientist in Generative 3D modeling at Salient (Acquired by Google Deepmind)
Part of a small, highly ambitious 4-person AI research team in weather forecasting, where we developed generative models for 3D for short- and medium-range weather forecasting. I led the development of multiple generations of model improvements across performance and efficiency. Subsequently acquired by Google Deepmind's weather team.
-
Sep 2021 — Apr 2026
Lecturer at ISM University of Management and Economics
Designed and taught four courses across the ML curriculum: Deep Learning, Elements of AI, Data Mining for Business Intelligence, and Econometrics.
-
2022 — 2026
Expert (Contract) at European Commission
Selected European Commission projects in deeptech, deep learning, and predictive analytics.
-
Aug 2022 — Feb 2024
Founder & CEO at Gauss
Bootstrapped an MLOps consultancy delivering predictive systems and machine learning infrastructure for small and mid-sized companies.
-
Feb 2020 — Jan 2022
Data Scientist at Luminor Group
Owned end-to-end data science workstreams in retail banking: business problem → model design → production deployment.
-
Jun 2019 — Aug 2022
Lecturer at Vilnius CODING School
Taught Python for data science, regression, classification, and clustering to small cohorts of ∼14.
Contact
Best by linkedin
Say hello.
Best way to reach me right now is through LinkedIn. I read everything and reply when I can.
LinkedInalso on Google Scholar