SuperAPI AI Directory
Partner with us中文AI newsFavorites

All AI tools

35,738 results · Page 735 of 993

LLMs-Universal-Life-Science-and-Clinical-Skills-

AI developer platforms

This project is a collection of reproducible, auditable research workflows for life sciences and clinical AI, maintained by Dr. MD BABU MIA. The repository transforms biomedical AI from scattered prompts into structured units of work, containing 1,207 skills across 54 categories, organized as SKILL.md files, reference notes, runtime adapters, and evaluation resources. Its core philosophy is that biomedical agents should understand their own domain, cite evidence, use tools deliberately, preserve provenance, and stop at appropriate human review boundaries. The repository supports translational research, clinical informatics, genomics, spatial and single-cell analysis, digital pathology, drug discovery, research automation, and modern agent infrastructure. The runtime layer BioKernel can route natural language research questions to appropriate skills, compose multi-step workflows, and evaluate outputs using biomedical standards. The content covers 3D genomics Hi-C analysis, AI provider integrations (OpenAI, Anthropic, Google, AWS, Azure, etc.), agent evaluation and observability, scientific discovery agents, clinical workflow automation, and more. Open-source components retain their respective licenses, while original architecture and workflow designs require attribution and a link to the original repository.

Read more

LLMs应用构建一年之心得

Other AI tools

This is a Chinese technical article jointly written by multiple large language model (LLM) application practitioners and translated by a translator, titled "One Year of Insights on Building LLM Applications (Part One)". The article was published in 2024 and summarizes the tactical experience gained over the past year in building LLM applications in real products, covering core areas such as prompt engineering, information retrieval and RAG, workflow optimization and fine-tuning, and evaluation and monitoring. The article points out that although the barrier to entry for building applications with LLMs has been lowered, creating effective products that go beyond demo effects remains challenging. The authors share specific practices such as multi-turn examples, chain-of-thought, structured inputs and outputs, keeping prompts concise, hybrid retrieval of RAG and keywords, prioritizing RAG over fine-tuning to acquire new knowledge, deterministic workflows, caching, assertion-based unit tests, and LLM-as-Judge evaluation, and remind readers that long-context models will not make RAG obsolete, that overemphasizing certain evaluations may harm overall performance, and that hallucinations are a stubborn problem. The article also provides background notes on the translation and a link to the original text, making it suitable for developers with some experience to quickly locate sections of interest for reading.

Read more

LLM训练-pretrain

AI learning

Historical third-party summary: Reprints on the Volcengine developer community and Machine Learning POD explicitly link Zhihu article 718354385 and credit ybq. The article discusses large-model pretraining workflow, data preparation and evaluation. A separately traceable resource-list index labels the same original URL LLM training/pretrain; source full text not verified for that index. Checked on 2026-10-18. The reprints are copies of one article, not independent studies, and do not verify author permission, technical effectiveness or current training results.

Read more