Profile banner
Towaki Takikawa

Towaki Takikawa

Co-founder & CEO at Outerport, ex-NVIDIA Research Scientist

San Francisco, California
Joined July 2024

Network

7.5K connections
πŸ’»
NVIDIA Researchers & Engineers
🧠
AI/ML/Robotics Researchers
πŸš€
Startup Founders
βš™οΈ
General Tech Engineers
πŸ€–
AI & Robotics Engineers
🏒
NVIDIA Pros
πŸ”¬
NVIDIA Rsch Sci

Summary

Towaki Takikawa is a deep expert in 3D generative AI, computer vision, and computer graphics, with a focus on neural fields and real-time rendering. His extensive research at NVIDIA and the University of Toronto has resulted in numerous highly-cited publications (over 4000 citations) in top-tier conferences like CVPR and SIGGRAPH, including foundational work on Magic3D, Gated-SCNN, and Neural Geometric Level of Detail. github+4
As the Co-Founder and CEO of Outerport (a YC S24 startup), Towaki is applying his expertise in AI and computer vision to extract structured data from complex technical documents and diagrams, such as CAD and P&ID. His work at Outerport aims to enable AI agents to understand visual information in engineering and other technical domains. github+3
Towaki has a strong academic background, holding a B.S. in Computer Science with Distinction from the University of Waterloo and pursuing a Ph.D. in Computer Science at the University of Toronto (currently on leave). His academic advisors include prominent figures in the field like Sanja Fidler, Alec Jacobson, Yuri Boykov, and Morgan McGuire. github+2
Beyond his research, Towaki has contributed to open-source software by creating Kaolin Wisp, a PyTorch library for neural fields research, and acting as a developer for Kaolin, a PyTorch library for 3D deep learning. github+1

Work

Education

Projects

Writing

Magic3D: High-Resolution Text-to-3D Content Creation

January 1, 2023

A foundational paper on scaling up 3D generation with smarter data structures for high-resolution text-to-3D content creation.

Favicon imagetovacinni.github.io

ATT3D: Amortized Text-To-3D Object Synthesis

January 1, 2023

Developed a framework for amortized text-to-3D object synthesis, allowing knowledge sharing between prompts for faster convergence and smoother interpolations in 3D asset generation.

Favicon imagetovacinni.github.io

Compact Neural Graphics Primitives with Learned Hash Probing

January 1, 2023

Introduced a method combining learned hash probing with hash tables for more compression and faster processing of neural graphics primitives.

Favicon imagetovacinni.github.io

Variable Bitrate Neural Fields

January 1, 2022

Explored learned indexing for more compressive neural field representations than hash tables and grids, enabling significant memory reduction and multiresolution representation.

Favicon imagetovacinni.github.io

RTMV: A Ray-Traced Multi-View Synthetic Dataset for Novel View Synthesis

January 1, 2022

Co-authored a paper on RTMV, a ray-traced multi-view synthetic dataset for novel view synthesis.

Favicon imagetovacinni.github.io

Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes

January 1, 2021

Proposed differentiable data structures, neural fields, and GPU programming for fast, real-time rendering of implicit 3D shapes, achieving high fidelity and efficiency.

Favicon imagetovacinni.github.io

Gated-SCNN: Gated Shape CNNs for Semantic Segmentation

January 1, 2019

Introduced a two-stream CNN architecture for semantic segmentation, explicitly wiring shape information for better computer vision by paying attention to object boundaries.

Favicon imagetovacinni.github.io

Hobbies

He was a Team Captain for FIRST Robotics Team 955, participating in robotics competitions. youtube

Notes

RCRachel Chalmerssaid they've seen
TTTowaki Takikawa
make standout contributions to autonomous driving projects that use AI.

Oct 22, 2025