Role: Founder & Lead Engineer
Duration: 2017–2019
Stack: React, Node.js, Express.js, Three.js, MongoDB, AWS S3, Blender Addon
The Problem: Selling Made-to-Measure Without Visualization
Independent opticians wanted to sell made-to-measure frames but lacked digital tools to make it work. Customers couldn’t preview customizations before ordering, forcing multiple consultations and risking dissatisfaction.
The challenge:
- Made-to-measure needed multiple in-person visits for fittings
- No real-time visualization of parametric adjustments (bridge width, lens height, temple length)
- Custom orders took 3+ weeks with no digital validation
- Without preview tools, bespoke sales meant higher return risk
Tartare fixed this: a CMS letting opticians sell made-to-measure frames with real-time 3D customization.
Tartare: CMS for 3D Customization
Tartare was a white-label platform that let eyewear brands deploy production-ready 3D configurators without custom development. Independent opticians could now sell parametric customization directly to consumers.
Core Platform Capabilities
For Brand Managers:
- No-code configuration interface - Upload frame models, define parameters, set constraints, publish
- Parametric frame morphing - Bridge width, lens height/width, temple length, rim width adjustments
- Color & material editor - Apply finishes (matte, glossy, metallic) to any frame surface
- Inventory management - Sync available sizes, colors, materials, constraints
- Order management - Track configurations through production pipeline
For Customers:
- Interactive 3D visualization - Real-time preview with WebGL rendering
- Parametric customization - Adjust measurements with instant 3D feedback
- Material/color selection - Visual material library with texture preview
Technical Stack
Frontend:
- React + TypeScript for admin interface and customer configurator
- Three.js + WebGL for real-time 3D rendering
- Parametric geometry engine for frame morphing
Backend:
- Node.js + Express.js API for configuration and order management
- MongoDB for storing brand settings, frame data, customer designs
- AWS S3 for model and texture asset storage
3D Pipeline:
- Blender Addon - Accelerated 3D model creation from CAD drawings
- Automated model optimization and texture baking
- Export pipeline generating production-ready assets
The Blender Addon: Bridging CAD to WebGL
One major problem: converting 2D manufacturing drawings into usable 3D web models. Manual modeling was slow and expensive. Brands needed a way to quickly convert their CAD libraries.
I built a Blender addon that semi-automated 3D model creation from vector drawings:
- SVG/CAD import parsing frame outlines
- Parametric mesh generation creating 3D geometry from 2D profiles
- Thickness & curvature application adding optical and structural form
- Material baking - Pre-compute realistic materials for WebGL performance
- Exporting optimizing models for web
The addon cut modeling from hours to minutes, letting brands rapidly onboard new frame collections.
White-Label Deployment at Scale
Tartare powered 5 premium eyewear brands with completely custom configurators deployed across their e-commerce platforms. Small in count, but the platform supported thousands of customers personalizing frames online.
Deployment Model
- Initial Setup - Brand provides 2D CAD drawings and design specs
- 3D Asset Creation - Blender addon accelerates model generation + manual refinement
- Configurator Configuration - Brand manager defines parameters, constraints, pricing
- White-Label Deployment - Tartare publishes to brand’s domain with custom branding
- Production Integration - Configurations automatically flow to manufacturing systems
Customization Depth
Beyond the core platform, I built extensive custom modules for each brand:
- Advanced AR try-on - Virtual Try-On using facial landmarks and device camera
- 3D facial scan integration - Manual frame positioning based on 3D scan (precursor to Heru)
- Parametric nose molds - Automated frame geometry adjustment for perfect fit around nasal bridge
- B2B wholesale tools - Configurators for business customers and resellers
- Custom checkout flows - Brand-specific order customization and upsell workflows
This level of customization explained why plugin architecture had limits - each client needed fundamental UI/UX tweaks that meant forking from core.
Key Technical Challenge: Extensibility vs. Stability
The Plugin System Attempt
I built a plugin architecture letting clients extend core functionality without forking:
- Plugin hooks throughout the application lifecycle
- Component composition enabling custom UI overlays
- Custom parameter types and validation rules
What worked:
- Minor customizations (color schemes, form fields, email templates)
- Additive features (new visualization modes, analytics integrations)
What broke:
- Fundamental UI/UX changes needing layout restructuring
- Core business logic modifications (pricing models, parameter constraints, order flows)
The Pragmatic Compromise
Rather than fight the architecture, I accepted controlled code forking for major clients:
- Maintain unified core library
- Branch versions for clients needing deep customization
- Periodic cherry-picking of critical fixes back to branches
- Accept higher maintenance burden in exchange for client satisfaction
This wasn’t elegant, but it worked in practice. Tartare stayed production-ready across all deployments while allowing genuine client differentiation.
Real-World Example: Heru (Cubitts)
Tartare’s most successful deployment became Heru, Cubitts’ 3D facial scanning + frame recommendation app.
While Heru evolved Tartare almost beyond recognition (rewriting ~100% of code), the architectural foundations stayed the same:
- Configuration-driven parameter system
- Blender-based asset pipeline
- React + Three.js technology stack
- Order management integration
Today, Heru powers made-to-measure eyewear sales across Cubitts’ 20 global stores. Tartare provided the scaffolding; Heru built the innovation on top.
See Heru case study for technical details.
Projects Using Tartare
- Heru (2019-2022): 3D facial scanning and eyewear recommendation for Cubitts
- Frame Up (2023-2024): Automated eyewear modeling from SVG to 3D, solving Tartare’s manual modeling bottleneck
- Heru 2 (2024-2025): Next iteration of Heru for professional optician consultations
Building something like this? Reach out or connect on LinkedIn .