01 · Roasts
The invisible launch
Eight named projects and 0 total stars: the shipping velocity is real, but discoverability has not boarded the train.
CI roulette
rover-notes and MBTALive bring serious automation; Edge-split and mbtalive-flutter still ask production to trust vibes.
Private-work fog
102 public commits look intermittent, while privateWorkLikely says the visible graph is only the trailer, not the feature film.
Portfolio beats audience
Five named upstream contributions are a stronger community story than 0 followers suggests—surface that work where people can verify it.
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% weight65C
- Consistency20% weight55D
- Quality20% weight82A
- Depth15% weight65C
- Breadth10% weight80A
- Community10% weight30F
03 · Stats
365-day commit heatmap
91 active days
Language distribution
- Java40%
- Jupyter Notebook28%
- Python7%
- TypeScript7%
- Swift6%
- Go4%
- Other8%
04 · Numbers
Owned repos
non-fork
11
Commits
last 12 months
102
Followers
0
Joined GitHub
Jul 2025
05 · Top repos
AMR5210 /
rover-notes
Rover Notes is a deeply engineered, documented Java/Spring hybrid-retrieval knowledge base with MCP, citations, Python ML services, browser coverage, production deployment templates, and measured retrieval-quality gates, but currently has no stars or forks.
AMR5210 /
MBTALive
A substantial, documented MBTA platform spanning Swift iOS features, Go services, Python notifications, Kafka, Terraform, and Kubernetes, with strong testing and deployment artifacts but no demonstrated adoption yet.
AMR5210 /
watchgpt
A documented, typed Apple Watch/Go AI application with production-oriented auth, streaming, caching, resilience, Kubernetes deployment, observability, and focused handler tests, but no demonstrated adoption or CI.
AMR5210 /
arm-create-mobileai
A substantial, well-documented QAT-to-GGUF research pipeline with physical iPhone benchmarks, custom Q2_K export, and Swift harnesses, but currently has no stars, tests, CI, or typed Python adoption signals.
AMR5210 /
Edge-split
Ambitious, well-documented EdgeSplit prototype implementing laptop-GPU prefill to Android decode with a patched llama.cpp state handoff, portable TCP framing, benchmarking, and a dashboard, but with no demonstrated adoption or sustained history.
AMR5210 /
mbtalive-flutter
A documented Flutter MBTA tracker with substantial service, state-management, commute-planning, persistence, and UI code, but no tests, CI, license, or demonstrated adoption.
AMR5210 /
AMR5210
A polished GitHub profile README presenting four substantial projects and five notable upstream contributions, but this repository itself is a small, non-code portfolio with no tests, CI, license, or fetched source files.
AMR5210 /
orbital-threat-assessment
Documented academic asteroid-PHA classification notebook with a substantial SMOTE/model-comparison workflow, but no visible adoption, tests, CI, licensing, or reusable typed package structure.
06 · Timeline
- Jul 20, 2025Joined GitHub
- Jan 17, 2026Created rover-notes — A personal knowledge base with hybrid retrieval and checkable citations, exposed over MCP for any MCP-capable agent to query.
- Feb 13, 2026Created mbtalive-flutter — Cross-platform Flutter app for real-time Boston MBTA subway tracking and smart commute planning.
- Feb 16, 2026Created MBTALive — Real-time Boston MBTA tracking, live Home Screen widgets, and smart transit notifications for iOS.
- Jun 29, 2026Created AMR5210
- Jul 2, 2026Created orbital-threat-assessment — ML pipeline for PHA detection: SMOTE + stratified 5-fold CV across LR, DT, RF, and LinearSVC. Best model reached 0.988 PR-AUC on a 449:1 imbalanced test set.
- Jul 9, 2026Created watchgpt — AI-powered Apple Watch app with a production grade Go backend. Cognito auth, Redis caching, SSE streaming, circuit breaking, and full observability on EKS.
- Jul 23, 2026Created arm-create-mobileai — Quantization-aware training for usable 2-bit Qwen3-0.6B inference on Arm mobile devices, with direct Q2_K GGUF export and physical iPhone benchmarks.
- Aug 3, 2026Created Edge-split — Disaggregated LLM inference: prefill on a laptop CUDA GPU, decode on an Android phone, with a checksummed TCP protocol handing off raw sequence state over Wi-Fi. TTFT down 65–82%.
- Sep 5, 2026Most recent push to AMR5210
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.