Time-lapse video of robotic arm printing PLA

RE.TAB BY ARM

Furniture without authorship. A log becomes prompt, code becomes craft.

The goal of this project is to reconfigure furniture-making as a feedback loop between nature, neural networks, and robotic tooling.
Beginning with found wood—each piece scanned, irregular, and full of material memory—we used image-to-3D AI models in ComfyUI to generate speculative table designs. These forms weren’t modeled—they were hallucinated, reconstructed from grain, void, and texture.

With robotic arms as collaborators, the workflow translated these generated forms into physical components. One robot carved the raw timber using wire-cutting strategies, while another extruded LX175 PLA as a secondary layer—wrapping the wood with synthetic geometry, completing the hybrid tectonic logic. Fusion 360 was used to simulate structural performance, while Grasshopper drove parametric variations and robotic motion control.

The result is not just a table—it’s a re-tabulation of material, machine, and memory.
It proposes that furniture design can emerge not from authorship, but from negotiation—between scanned matter, AI suggestion, and robotic execution.

The Challenge: In a market saturated with mass-produced goods, design-conscious consumers are seeking unique, sustainable furniture with a compelling story. Meanwhile, valuable resources are wasted when irregular or “flawed” timber is discarded by conventional manufacturing.

My Role & Approach: I led the end-to-end process as the sole Designer, AI Trainer, and Robotic Fabricator. My approach was to create a sustainable manufacturing model where design authority is shared between natural materials, generative AI, and robotic execution.

Research & Insights: My research into material lifecycles and sustainable design revealed a key insight: the irregularities in reclaimed wood are not defects, but unique “design prompts.” I hypothesized that AI could act as a “translator,” converting this unstructured natural language into a precise geometric language for robotic manufacturing.

Design & Prototyping: I developed a novel digital-to-physical workflow: 1) Digitization: 3D scanning reclaimed wood to capture its unique geometry. 2) AI Generation: Using an image-to-3D model in ComfyUI to “hallucinate” table designs that organically responded to the wood’s form. 3) Robotic Collaboration: Orchestrating a 6-axis robotic arm to carve the timber while another extruded PLA for a hybrid structural skin.

Deciding other detailed tabletop shapes via ComfyUI

Time-lapse of robotic arm cutting wood

Testing & Iteration: Before fabrication, all AI-generated designs were validated through structural stress analysis in Fusion 360 to ensure real-world stability. The robotic toolpaths were iteratively simulated in Grasshopper to maximize material efficiency and precision.

Outcome & Impact: The result is a one-of-a-kind furniture piece that is a physical record of its own creation. This project demonstrates a new paradigm for automated, hyper-personalized production. Its core principle of upcycling reclaimed materials aligns directly with the corporate sustainability goals championed by Google, showcasing how to turn waste into high-value products.

Technologies Used

Toolchain:

Material Digitization: 3D Scanning · Mesh Cleanup
Generative AI: ComfyUI · Stable Diffusion (image-to-3D prompt)
Parametric Design: Grasshopper · Fusion 360
Robotic Fabrication: 6-axis Robotic Arm · Wire Cutting · PLA Extrusion (LX175)
Structural Simulation: Fusion 360 (stress analysis)
Motion Control & Scripting: Grasshopper · Python · Robot Plugin
Prototyping Tools: PLA Filament · CNC Cut Timber
Software Ecosystem: Rhino · Fusion 360 · GitHub

→ These tools were choreographed into a hybrid digital-physical workflow where material memory, AI inference, and robotic execution converge—automating the design and making of custom furniture directly from site-found wood.

Grasshopper computing robotic arm wood-cutting path

Robotic arm printing PLA

Strategic Takeaways: Generative Design: Leveraging AI to create novel solutions from non-standard, real-world inputs. Human-Computer Collaboration: Designing workflows that seamlessly integrate human creativity, AI inference, and robotic precision. Digital-to-Physical Prototyping: Mastering the full-stack process from virtual simulation to tangible, functional output.

Next Steps

Interested in learning more about this project or discussing potential collaborations? Feel free to contact me or explore my other projects.