01 · Roasts
The Solo Hermit
soloPct=100 across every single repo. Not one PR, not one co-author, not one issue from a stranger. Your GitHub is a beautifully furnished room that nobody else has ever entered.
Burst Mode Developer
DM2026-Assignment-3: 30 commits in ~1 month. visual_recognition-fp: 30 commits in 8 days. cvat-sam3: 9 commits in 3 days. Your entire commit history reads like a finals week panic schedule, not an engineering career.
2 Stars, 23 Repos
You've shipped 9+ projects with typed configs, ARCHITECTURE.md files, and RFC-compliant canonical JSON — and the entire internet has rewarded you with 2 stars. Marketing: consider it.
CI Avoidance Champion
pi-workflow, DM2026-Assignment-3, DM2026-Final-Project, visual_recognition-hw4, and visual_recognition-fp all have HAS_TESTS=yes but HAS_CI=no. You write tests but won't let a machine run them. Trust issues?
Over-Engineered Coursework
DM2026-Assignment-3 ships with a SUBMISSIONS.md tracking 42+ feature aggregations, a PDF report, and 5 test files — for a university assignment. The professor asked for a notebook, not a production ML system.
Built using
Zoral
Shadows one worker for a week, then takes over their job with zero extra setup. Behaves exactly like the original.
zoral.ai
02 · Category breakdown
- Impact25% weight62C
- Consistency20% weight65C
- Quality20% weight72B
- Depth15% weight65C
- Breadth10% weight65C
- Community10% weight55D
03 · Stats
365-day commit heatmap
22 active days
Language distribution
- TypeScript36%
- Python36%
- Jupyter Notebook12%
- JavaScript5%
- Shell4%
- Dockerfile1%
- Other6%
04 · Numbers
Owned repos
non-fork
19
Commits
last 12 months
47
Followers
11
Joined GitHub
Mar 2020
05 · Top repos
RayhanHaqi /
pi-workflow
TypeScript bounded agentic-coding workflow kernel with deterministic M1–M5 foundations, comprehensive schema validation, immutable state-machine semantics, and secure filesystem confinement. Shipped with tests, CI absent but typed with strict tsconfig.
RayhanHaqi /
DM2026-Assignment-3
University assignment repo for human activity recognition via accelerometer data. Achieves 0.7897 public score on Kaggle using TabPFN V3 with calibration. Well-structured with 11k+ labeled training files, temporal feature engineering, and reproducible scripts, but limited reusability outside coursework context.
RayhanHaqi /
DM2026-Final-Project
Course final project for natural disaster severity prediction (0–5 scale) using XGBoost and ordinal classification. Includes temporal feature engineering, probability caching, and ensemble blending. Documented with README, design artifacts, and comprehensive unit tests. Typed Python with structured src/ layout, but lim
RayhanHaqi /
visual_recognition-hw4
Course homework project for image restoration using PromptIR model on rain/snow degradation (3,200 training pairs). Ships with typed Python, comprehensive training/inference pipeline, datasets, and unit tests. No external impact beyond coursework.
RayhanHaqi /
visual_recognition-fp
NYCU Visual Recognition course final project: Kaggle sea-lion population counting competition (RMSE 14.44). Typed Python with tiled inference, ensemble pipeline, tests, and structured src layout. Personal course submission without external adoption signals.
RayhanHaqi /
composer-deepswe-estimation
Reproducible cross-benchmark linking pipeline for estimating Composer 2.5 DeepSWE performance using 14 overlapping model-effort pairs; Python typed workflow with CI, tests, and methods comparison.
RayhanHaqi /
composer-deepswe-trials
Single-purpose trial publication repo for Cursor Composer 2.5 on DeepSWE v1.1. Includes validation scripts, leaderboard charting (matplotlib), and JSON schema—well-structured experimental benchmark measurement with clear docs and CI, but limited scope and zero adoption.
RayhanHaqi /
cursor-codex-reviewer
Experimental Cursor skill (v0.2.0) for Codex-first planning with structured read-only investigation and approval gates. Shell-based, well-documented, and safety-focused but nascent (0 stars, 5-day lifespan).
RayhanHaqi /
cvat-sam3-auto_track
Fresh fork/variant of CVAT with SAM3 auto-tracking integration (89MB codebase, 9 commits in 3 days). Has docs and CI setup but minimal independent contribution depth and no original stars/adoption.
06 · Timeline
- Mar 14, 2020Joined GitHub
- May 4, 2026Created DM2026-Assignment-3
- May 15, 2026Created visual_recognition-hw4
- May 15, 2026Created DM2026-Final-Project
- May 23, 2026Created visual_recognition-fp
- Jun 16, 2026Created composer-deepswe-estimation
- Jun 20, 2026Created composer-deepswe-trials
- Jun 21, 2026Created cursor-codex-reviewer
- Jul 1, 2026Created cvat-sam3-auto_track
- Jul 29, 2026Created pi-workflow
- Aug 27, 2026Most recent push to pi-workflow
07 · Compare
08 · Rubric
How this score was produced
Overall = Σ (category × weight) + gentle top-end curve
Tier thresholds
▸ How the pipeline works
- 01Scrape.Pull every non-fork repo pushed in the last 90 days, plus your contribution calendar, followers, and language byte counts — straight from GitHub's REST & GraphQL APIs.
- 02Triage.A small model reads every repo's file tree + README and picks the 20 files per repo that actually reveal how you code.
- 03Grade each repo. All repos run in parallel through a fast scoring model that reads the picked files and rates each one independently on Impact, Quality, and Depth — with evidence citations.
- 04Aggregate. A larger reasoning model combines the per-repo scores with server-computed stats (heatmap, commit cadence, language entropy, follower count) to produce the 6-dimension profile score + roasts.
- 05Correct.Deterministic server-side checks enforce anchor-scale floors (e.g. a profile with 2,000+ public commits can't score 30 Consistency) and recompute the final verdict.
~90 seconds per profile, ~$0.25 in compute. Total of ~240 files read across your top-12 repos. One rating per GitHub account per day.
▸ Data sources & caveats
- Heatmap & commit totals: GitHub GraphQL
contributionsCollection— covers the last 365 days, includes private repos when the user has opted in (default). - Language %: byte totals across the top 30 owned non-fork repos.
- Curve: a small upward nudge centered on raw score ≈ 70, capping at 100. Prevents specialists from being unfairly penalised for narrow breadth.
- Anchor corrections: when server-measured signals (e.g. privateWorkLikely, multiRepoVolume, follower count) mandate a minimum category score, the aggregation step enforces it. These are signal-conditional, not identity-based floors.