Hi, I’m Jakub (Kuba) Strawa 👨🏻💻
I'm an AI Researcher and Research Engineer specializing in LLM training, post-training, multimodal models, and native multimodality. My work bridges the gap between cutting-edge AI research and scalable, real-world applications. Currently, I work at Stonly, where I focus on developing, training, and rigorously evaluating AI Agents (Agents & Evals).
Prior to Stonly, I was an AI Researcher at TCL working on Multimodal Reasoning, and I’ve spent time in R&D at Raiffeisen Bank and Roche developing custom multimodal solutions. Throughout my career, I’ve built and deployed enterprise-grade AI applications for Fortune 500 companies, worked alongside top-tier AI researchers in China, and collaborated directly with the Qwen team on multimodal post-training. In my spare time I share my knowledge in technical posts about AI on my blog.
Upcoming talks
- Native Multimodality: Beyond Language-Centric Multimodal Models — Data Science Summit Warsaw, Summer Edition (19.06.2026)
- Native Multimodality: Beyond Language-Centric Multimodal Models — GenAI Kraków #28 (29.06.2026)
Research interests
- Multimodal models (vision, language, audio) & native multimodality
- Multimodal synthetic data
- Reinforcement learning
- Agents & evals
Latest publication
- Improving KAN with CDF normalization to quantiles , J. Strawa, J. Duda
Elsewhere