Free - Flow!
Free - Flow!
A Flow Matching model for SVG generation
(Under development)
Embedded Files
Spick and Span!
Spick and Span!
A VLA model for long horizon Home Rearrangement Tasks
(Under development)
Brick-by-Brick
Brick-by-Brick
A Physics informed VLA model for brick structure generation with grounded reasoning
"The Ghosts" of CAD
"The Ghosts" of CAD
Conversational Agentic AI for CAD generation
Photo2Fab
Photo2Fab
A Vision model for branch skeleton detection
Split_pts
Split_pts
Panoptic Point cloud segmentation model
Spawn Out Loud!
Spawn Out Loud!
Text conditioned SDF Diffusion model
Nexus Events
Nexus Events
Non-player character trained with RL for a game based on chain reaction
Gen3D Repertoire
Gen3D Repertoire
Experiments with 3d model generation dealing with Point clouds, voxels, meshes, nerfs, SDFs, Gaussian splats.
Ugly design
Ugly design
Image based Diffusion model for Residential floor plan generation
3T's model
3T's model
DiT model for Sketch to 3d form generation
Inverse Chopsticks
Inverse Chopsticks
MCP for Real time interaction with robotic arms using VLMs
IN and OUT
IN and OUT
A Brick construction simulator with motion capture
Tweak Ai
Tweak Ai
A GAN model to predict the buildable volume with a given context
Bunch of Bargains!
Bunch of Bargains!
A Multi agent simulation system for Supply Chain
negotiation
Cybernetic Flappy bird
Cybernetic Flappy bird
A Flappy bird game that adapts as the game continous with respect to the players actions
Spider
Spider
A Parametric grasshopper plugin that aids designers to analyze , visualize data and generate urban layouts with Rhino-Grasshopper
Generative Parkng
Generative Parkng
A parametric tool that generates open parking layouts based on the given context
I'm a generative AI researcher and design technologist working where machine learning meets spatial reasoning and human creativity. I build end-to-end systems, from ML pipelines and model training to the interfaces around them, that turn complex design and planning problems into usable tools. My work spans VLM and diffusion-driven co-pilots, reinforcement-learning agents, 3D generative models, and multimodal workflows connecting geometry, simulation, and evaluation.
I care as much about usefulness as novelty. Human-in-the-loop feedback, custom evaluators, clear metrics, and reproducible experiments are built into how I work, and I'm equally at home prototyping inside design toolchains and shipping production-ready services. That ranges from a published Grasshopper plugin used by designers to fine-tuned domain language models evaluated against real tasks.
At Carnegie Mellon and Studio.Compute, I've led cross-disciplinary teams, mentored fellow researchers, and built interactive tools for automation, decision support, and immersive XR. Whether the work is augmenting creativity, spatial planning, or 3D generation, the goal is the same: augment cognitive process and accelerate high-quality outcomes.
-Karthick Raja BG
+1(412)-689-6346 | karthickrajaforjobs@gmail.com | linkedin | CV
Page updated
Google Sites
Report abuse