Skip to content
View edenfunf's full-sized avatar

Highlights

  • Pro

Organizations

@opensource4you

Block or report edenfunf

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

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

Report abuse
edenfunf/README.md

Hi, I'm Eden πŸ‘‹

18-year-old developer from Taiwan πŸ‡ΉπŸ‡Ό who loves open source.

πŸ€– AI agents & harness engineering: multi-agent orchestration for long-running coding agents
πŸ”Œ Embedded systems: RISC-V, Edge AI, mmWave radar, ESP32 and Raspberry Pi

πŸ₯‡ Gold, Andes RISC-V Contest Β Β·Β  πŸ₯‡ Gold, All Ring Tech Innovation Contest

C++ C CUDA Python TypeScript JavaScript

πŸš€ Things I built

Highlight Project
πŸ”₯ #1 on GitPop
JavaScript weekly trending
ReelMimic: a multi-agent harness that turns Claude Code or Codex into a video crew. Show it a video you love: up to 6 agents build a new one in the same style, and a fresh agent reviews every shot, so no agent grades its own work.
πŸ† Nantou track winner
MODA Civic Action Hub 2026
AidFlow: disaster reporting platform built with Nantou County Government. Citizens report without logging in, nearby reports merge into cases, and county staff triage them. Demo site
πŸ† Merit Award
DIGITIMES GenAI Hackathon 2026
Family Avatar: grandparents speak Taiwanese, and an AI family member answers on video in ~2 s. Runs end to end on one A10G GPU.
πŸ† Winning Team
MODA Civic Tech Challenge
ResQLink: describe a disaster in one sentence; an AI agent picks from 72 modules and generates the whole relief system.
πŸ“ˆ 1.72Γ— tokens/s
βˆ’42% cost per token
TuneForge: tunes LLM inference on Arm64 cloud. An AI agent reads hardware counters and proposes the next experiment; the harness changes one setting at a time, measures it, and rejects anything that hurts model quality.
πŸ“¦ pip install promptc promptc: grades your Claude Code setup (A–F) for duplicated skills and wasted context.

🌍 Open source

Project Merged PRs Highlights
 OpenClaw ⭐ 394k 81 Top external contributor to the LINE channel, ~45% of its code
 Microsoft APM ⭐ 4.2k 48 #1 external contributor, fixed an enterprise credential leak (#1459)
Β NVIDIA CCCL Β· TensorRT-LLM 15 Single-pass is_partitioned, one GPU kernel launch instead of two, up to 2Γ— faster (#8427)
Β llama.cpp Β· Β RAGFlow Β· Β openai-python Β· more 5 CUDA kernel fix in llama.cpp, RAG and SDK fixes

β†’ See all merged PRs

All merged PRs by repo (auto-updated weekly)

Open Source Contributions

πŸ“« Contact

X / Twitter Β· fya973619@gmail.com

Pinned Loading

  1. openclaw/openclaw openclaw/openclaw Public

    The AI that really does things. Any OS. Any Platform. The lobster way. 🦞

    TypeScript 394k 82.8k

  2. microsoft/apm microsoft/apm Public

    Agent Package Manager

    Python 4.2k 417

  3. reelmimic reelmimic Public

    Show it a video you love. Get a new video in the same style. An AI crew (Claude Code or Codex) plans, builds and reviews it with you.

    JavaScript 1.8k 55