Roman Sitelew

I have authored dozens of research papers, including on evaluating AI self-awareness and "sideloading" - the methodology of building highly accurate digital replicas of real people using LLMs. My code and benchmarks are trusted by industry leaders, as evidenced by my contributions to Google's open-source projects, including BIG-bench and the Agent2Agent protocol.

### KEY PAPERS

* Turchin, Sitelew. Sideloading: creating a model of a person via LLM with very large prompt (2025)
* Dhole, et al. Nl-augmenter: a framework for task-sensitive natural language augmentation (2023)
* Srivastava, et al. Beyond the imitation game: quantifying and extrapolating the capabilities of language models (2023)
* Sitelew. Simple computer modeling of biological processes on the example of "Petri" app (2008)

---

### SELECTED CONTRIBUTIONS TO OPEN SOURCE PROJECTS

* 2025 | Google's Agent2Agent (A2A) protocol
* https://github.com/a2aproject/a2a-samples/tree/main/samples/python/agents/number_guessing_game
* 2020 - 2021 | Google's "BIG-bench" LLM benchmark
* https://github.com/google/BIG-bench
* Tasks & datasets: self_awareness, simple_ethical_questions, simple_text_editing, misconceptions_russian, cryobiology_spanish

---

### LANGUAGES

* Conversational: English, German, Russian. Learning: Chinese. After that: Spanish

---

### HOBBIES

* Jogging, learning languages, personal automation, robot maintenance