Skip to content
View Picrew's full-sized avatar
👋
👋

Highlights

  • Pro

Block or report Picrew

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Picrew/README.md

Hi there 👋

Typing intro

I am Junjie Li, an M.S. student at the Chinese Academy of Sciences (2024-2027).

I am currently an LLM Algorithm Intern at Apple (May 2026 - Present).

I recently completed an LLM research internship at Microsoft (Apr 2025 - Apr 2026), and earlier internships at Shanghai AI Lab (Dec 2024 - Apr 2025) and NetEase (Aug 2024 - Dec 2024).

Those experiences shaped how I build: keep the core idea sharp, keep the system practical, and keep the iteration loop fast.

My main interests are long-context generation and reasoning, LLM evaluation, and agent systems. I enjoy turning research ideas into tools that are simple to run, easy to debug, and reliable in real workflows.

I share notes, prototypes, and experiments on my website: picrew.github.io.

If you'd like to discuss long-context LLMs, evaluation, agent systems, or other research and engineering ideas, feel free to reach me at lij850601@gmail.com.

Pinned Loading

  1. ConStory-Bench ConStory-Bench Public

    ConStory-Bench (ACL 26')

    Python 190 5

  2. awesome-agent-harness awesome-agent-harness Public

    An awesome list of Agent Harness engineering resources, including GitHub projects, tools, benchmarks, and practical guides.

    Python 1.4k 117

  3. awesome-llm-story-generation awesome-llm-story-generation Public

    A curated list of LLM papers and open-source projects for story, novel, and script generation.

    83 10

  4. OpenAI_Gradient_Atelier OpenAI_Gradient_Atelier Public

    Generate OpenAI-style diffusion and material gradient backgrounds in the browser.

    JavaScript 8 3

  5. Daily-blog Daily-blog Public

    Daily Chinese tech digest from Karpathy’s 90 curated blogs, with AI ranking, link analysis, and a static web reader. | 基于 Karpathy 精选 90 个博客的中文每日技术摘要,包含 AI 排序、链接分析和静态阅读站。

    TypeScript 39 3

  6. OpenRAG-Skill OpenRAG-Skill Public

    Prompt-only RAG with strict citations for in-context documents | 面向上下文内文档的纯 Prompt 引用型 RAG Skill

    31