Developer Profile

Daisuke Majima (MLBoy)

@john-rocky

Japan

GitHub

Impact Score Overview

Developer Impact & Statistics

Final Score

78 / 100Raw: 1138.00

Repository Score

76 / 100Raw: 312.00

PR Score

88 / 100Raw: 2200.00

Contribution Score

43 / 100Raw: 75.00

Score Weight Distribution

finalScore = repoScore * 0.45 + prScore * 0.45 + contributionScore * 0.10

Repository Score (45%)12% of final signal

Pull Request Score (45%)87% of final signal

Contribution Score (10%)1% of final signal

Top Work

Top Work & Highlights

Top Repositories

Most impactful repositories and pull requests

#1CoreML-Models

stars1859

forks172

Swift 8800%Python 1200%

Score

#2coreai-model-zoo

stars402

forks25

Python 6900%Swift 2600%Metal 400%

Score

#3CoreML-LLM

stars183

forks24

Python 6500%Swift 3200%Shell 200%

Score

Top Pull Requests

Most impactful repositories and pull requests

#1Add Apple Core AI export

in ultralytics/ultralytics

stars on the PR repository61199

+299/-41

Python 10000%

121

Score

#2Fix Zamba2 construction for num_mem_blocks > 1 checkpoints

in huggingface/transformers

stars on the PR repository164717

+43/-2

Python 10000%

101

Score

#3Release temp .mlpackage eagerly in MLModel.__del__

in apple/coremltools

stars on the PR repository5411

+162/-0

Python 8400%C++ 1400%Swift 100%

Score

Community Contributions

comments

#1Issue

Zamba2: `num_mem_blocks > 1` checkpoints cannot be constructed — `block_id` uses the global layer index while the tie cycle uses hybrid order

huggingface/transformers

stars164717

comments8

Score

#2Issue

`export_llm` Core ML builds fail at execute: "must be an MLState" even though `take_over_mutable_buffer=False`

pytorch/executorch

stars4979

comments6

Score

#3Issue

export_hf int4: the named int4 recipe (dynamic_wi4_afp32) is channelwise and collapses small LLMs — expose/recommend a blockwise int4 recipe

google-ai-edge/litert-torch

stars1091

comments9

Score

Scoring Transparency

▼

Repositories analyzed

Pull requests analyzed

118

Merged external PRs counted

118

Own-repo PRs ignored

Unmerged PRs ignored

Unique external repositories

Issues analyzed

100

External issues counted

100

Discussions analyzed

External discussions counted

How this score was calculated

View the full scoring methodology with formulas, penalties, and step-by-step flow.

View full methodology