Skip to content
View david-dabert's full-sized avatar

Block or report david-dabert

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
david-dabert/README.md

David Dabert

Systems Architect & Founder — Building Zero-Server Computer Vision, Biomechanical Kinematics & Sovereign Client-Side AI.

Twelve years coordinating 15-partner international consortiums across six countries (EU, FCDO, AFD) — now engineering high-performance, private, client-side tools that eliminate server dependencies.

LinkedIn Portfolio Newsletter


🚀 Flagship Project

WorkoutVision — Zero-Server On-Device Movement Intelligence

Real-time computer vision workout and biomechanical analysis running 100% inside the browser. No cloud, no video uploads, no accounts, zero backend costs.

  • Architecture: MediaPipe WASM & WebGPU pose detection in Web Workers with concurrent frame backpressure regulation.
  • Accuracy & Validation: Deterministic kinematic counting validated against academic benchmarks (Countix MAE 0.90) and exact paired McNemar statistical test harnesses ($p = 1.00$ regression-free invariant).
  • Thermal Duty Cycle: Dynamic frame throttling (30 fps active $\to$ 4 fps rest), slashing mobile SoC compute by >70% on iOS Safari.
  • Stack: TypeScript, React, MediaPipe BlazePose, Web Audio API, Web Workers, IndexedDB, TailwindCSS.

🛠️ Open-Source Research & Tooling

Specialized, zero-dependency client-side libraries extracted from production computer vision systems:

Repository Description Technologies
gravity-pose Monocular 3D pose rectification & gravity alignment for floor-tilted smartphone cameras ($15^\circ-35^\circ$ pitch). Restores 100% true vertical ROM. TypeScript, 3D Rotation Matrices, MediaPipe
kinematic-inpaint 3D rigid kinematic occlusion inpainting & bone-length trilateration. Solves 45 cm Olympic bumper plate & gym machinery joint dropouts ($dL/dt = 0$). TypeScript, Kalman Filter, Analytical Trilateration
acoustic-haptics Missing-fundamental psychoacoustic sub-bass (virtual 80 Hz) & noise-penetrating transient cue engine cutting through 90 dBA gym noise without earbud distortion. Web Audio API, Psychoacoustics, Sound Synthesis
sovereign-health Zero-server streaming parser for 1+ GB Apple Health export.xml and Google Takeout CSVs with sub-15 MB RAM footprint. TypeScript, Streaming Regex, Local-First

💡 What I Bring to the Table

  • Systems Engineering & Ground-Truth Discipline: Shaped by leading 15-partner international consortiums across 6 countries. I architect before I build and evaluate with mathematical rigor (bootstrap confidence intervals, empirical benchmark suites).
  • Zero-Server Client-Side Purity: Strong advocate for local-first computing—moving expensive cloud compute (computer vision, audio synthesis, health analytics) directly to edge devices via WASM and modern Web APIs.
  • Fast Polymathic Execution: Mastery of computer vision, WebCodecs, digital signal processing, biomechanical kinematics, and French typography conventions.
  • Bilingual Leadership: Bilingual French/English, five languages total. Editor of The Practitioner's Read (~1,100 institutional readers across European Commission, Quai d'Orsay, IOM, ETF, EIB).

🏛️ Background & Accreditation

  • Programme Coordination: Bioforce RNCP Level 7 (Humanitarian Programme Coordination)
  • Institutional Consortiums: 15-partner DFID-UKAID consortium coordination (Dakar 2018–2019)
  • European Commission: SALTO-YOUTH accredited trainer (17 domains)
  • International Leadership: 10 years delivering francophone programmes across London institutions (2008–2017)

📍 Bordeaux, France

Popular repositories Loading

  1. workout-vision workout-vision Public

    On-device AI workout analysis. MediaPipe pose detection, seek-based frame extraction, rep counting, form scoring, biomechanics. React 19 PWA, zero backend.

    JavaScript 3

  2. david-dabert.github.io david-dabert.github.io Public

    Personal site — programme coordination, grant operations, coalition building

    HTML

  3. portfolio portfolio Public

    Personal portfolio site

    HTML

  4. afd-grant-monitoring afd-grant-monitoring Public

    Grant portfolio monitoring toolkit. AFD, EU, multi-donor operations. Dashboard, RAG scoring, indicator tracking, automated report pre-reads.

    Python

  5. david-dabert david-dabert Public

    Profile README

  6. aivision-live-wall-vercel aivision-live-wall-vercel Public

    HTML