Demircan Tas

t a s d @ m i t . e d u
Google Scholar / GitHub / Twitter

About me

I am a PhD student in EECS at MIT, advised by Stefanie Mueller in the HCI Engineering Group at CSAIL. I work on generative AI for functional, fabricable 3D objects: letting people shape objects with their hands, speech and real-world references, and making sure the results work once they are printed.

Before MIT, I worked in industry as a technical artist and technical director, building graphics pipelines for architecture, industrial design and entertainment, and as a researcher at Istanbul Technical University on computer vision for architectural heritage. I hold an SMArchS in Design & Computation and an SM in EECS from MIT.

Research

Generative AI that makes objects that work in the physical world. Generated 3D shapes often look right but fail once they are made: openings are sealed, walls are too thin, parts don't connect and nothing moves. I build interfaces and methods that close this gap: repairing generated shapes, composing new objects from real ones in mixed reality, inferring physical conditions from how people describe an object's use, and building precise joints into generated shapes.

Earlier work (click to expand):

▸ Design & Computation: Parametric Paint-Over, my SMArchS thesis

Parametric and algorithmic design map parameters to narrow solution spaces that often exclude the changes made while iterating on drawings or physical prototypes. My thesis combined image and text conditioned diffusion encoders to sample control vertices for weighted Bezier splines (NURBS) in million-dimensional latent spaces, finding non-linear interpolations among 3D objects that mimic parametric transformations.
DSpace, Inverse Drawings, Pixel Rules

▸ Virtual Reality and HCI: work from my research at ESI

The Gorge VR project for MIT Climate Machine

Wolman Game for the MIT Urban Metabolism Group

▸ Construct(): speech and gesture to Minecraft blocks

I co-founded Construct(), a speech/gesture-to-blocks AI for games with building mechanics, through MIT Sandbox and DesignX.

DesignX pitch

Hands on (or off) demo

▸ Demoreel: projects from my studies at MIT

Highlighted papers (All papers)

Printed dragon mug repaired with InstructMesh

InstructMesh: Text-Guided Selective Refinements of Generative 3D Models for Fabrication
Faraz Faruqi, Ahmed Katary, Demircan Tas, Theresa Hradilak, Ning Zhang, Jiaji Li, Fabian Manhardt, Martin Nisser, Vrushank Phadnis, Ruofei Du, Federico Tombari, Megan Hofmann, Stefanie Mueller
UIST 2026.
[Paper]

Select the part of a generated 3D model that's wrong, say what to fix, and get a printable object.

MiXR: Harvesting and Recomposing Geometry from Real-World Objects for In-Situ 3D Design
Demircan Tas*, Faraz Faruqi*, Ning Zhang, Arthur Caetano, Dingning Cao, Niccolo G Meniconi, Oğuz Arslan, Misha Sra, Ruofei Du, Stefanie Mueller, Mustafa Doga Dogan
arXiv 2026.
[Paper]

Capture real objects in mixed reality, arrange their parts by hand, and generative AI fuses them into one printable object.

Supporting Design Exploration through In-Situ Guidance for 3D Modeling
SooYeon Ahn*, Demircan Tas*, In-Chang Baek, Stefanie Mueller, Jin-Hyuk Hong
UIST 2026 Poster.

* equal contribution

Teaching

6.8300/6.8301: Advances in Computer Vision (TA, Spring 2023, Spring 2024)
6.9020: How to Make (almost) Anything (TA, Fall 2022)
4.542: Background to Shape Grammars (TA, Spring 2022)
4.500: Introduction to Design Computing (TA, Fall 2021)
Lecturer at Anima School, Istanbul: 3D modeling, scripting and rendering (2014–2021)

My How to Make (almost) Anything pages (Fall 2021).