We provide RLHF services to tech bio companies looking to improve their LLM models. Some interesting projects we have worked on so far are -

• Breaking LLMs: Expertly crafted conceptual biology and biostatistics questions designed to challenge and enhance large language models (LLMs). Our questions were meticulously curated by specialists across multiple biological domains, including ecology, bioinformatics, cell biology, microbiology, and immunology, ensuring accuracy and relevance. Fine-tuned for maximum impact, these questions helped advance the capabilities of AI models in biology. These questions were chosen by their ability to break existing models by failing to provide correct answers. The content was advanced grad school biology curated by PhDs and advanced Master’s students.

• Teaching LLMs: In another project, we focused on teaching LLMs to understand specific topics in biology through a meta-simulation of a student-teacher dynamic. Starting from foundational concepts and progressing to graduate-level material, this deep-dive was structured within a 40-50 turn limit. Given this constraint, we concentrated on delivering a highly detailed exploration of each topic, ensuring depth and specificity.