Khushal Sethi

I build autonomous systems that design and improve other systems.

Intelligence compounds when friction is removed.

SCROLL
01 ABOUT
Khushal Sethi

I'm building DragonX Systems full time — autonomous infrastructure for chip design and the physical layer.

My work sits at the intersection of computation, capital allocation, and engineered intelligence.

I've worked on accelerator architectures and systolic array performance modeling. Previously I cofounded Kiwi AI. Today my focus is broader: building tools that collapse complexity and compress the distance between thought and execution.

Long-term, I'm interested in metamachines — systems that automate the creation of other machines across industries.

CURRENTLY

Founder · DragonX Systems

PREVIOUSLY

Cofounder & Head of Engineering · Kiwi AI
Founder · GodelAI

PREVIOUSLY

Lead Design Engineer T3
Tensilica, Cadence · US

LIFE GOAL

Automate all work and eliminate human suffering.

EDUCATION

MS Stanford University
B.Tech IIT Delhi · AIR 79 / 1.5M

02 CLIMATE

Stanford EE PhD · Chip designer · Gen Z

The world is burning. Not later. Now. And it is burning faster every year.

Heat stacks. Systems strain. Feedback loops feed themselves. That is exponential. Physics does not pause for comfort, elections, or old people saying they survived worse.

Older generations used the last stable decades and left the curve pointing up. Then they said the next generation would fix it. That was the rug pull.

This time there is no quiet after. If the physical stack goes — energy, food, water, machines, order — people do not get a sequel. They die. All of them. Rich, poor, young, old. The fire does not keep a guest list.

I did not study electrical engineering to decorate a funeral. I studied it because reality is circuits, power, heat, and limits. Those limits are arriving.

Do not tell me the world always works out. That was true only while surplus lasted. Surplus is ending.

See the curve. Say it plainly. The world is burning exponentially. If no one holds the physical layer, this ends with everyone dead.

Awareness is not a feeling. It is looking at that and not lying.

03 FOCUS
01

Chip Design Engines

Autonomous systems that translate high-level specifications into silicon-ready layouts. Intelligence at the physical layer.

02

Autonomous Workflows

Orchestrating complex multi-step processes that design, simulate, and manufacture — end-to-end, with minimal human-in-the-loop.

03

Intelligence Infrastructure

The intersection of code, capital, and computation. Building the substrate on which scalable intelligence runs.

04 PUBLICATIONS
ARXIV 2022

DRAGON (Differentiable Graph Execution): A suite of Hardware Simulation and Optimization tools for Modern AI/Non-AI Workloads

Khushal Sethi

PAPER
ARXIV 2022

Efficient On-Chip Communication for Parallel Graph-Analytics on Spatial Architectures

Khushal Sethi

PAPER
IEDM 2021

The future of hardware technologies for computing: N3XT 3D MOSAIC, illusion scaleup, co-design

RM Radway, K Sethi, W-C Chen, Jimin Kwon, S Liu, TF Wu, E Beigne, MM Shulaker, H-SP Wong, S Mitra

A-SSCC 2020

Compacc: Efficient hardware realization for processing compressed neural networks using accumulator arrays

Zexi Ji, Wanyeong Jung, Jongchan Woo, Khushal Sethi, Shih-Lien Lu, Anantha P Chandrakasan

ARXIV 2020

NV-Fogstore: Device-aware hybrid caching in fog computing environments

Khushal Sethi, Manan Suri

JETC 2020

GPUOPT: Power-efficient photonic network-on-chip for a scalable GPU

Janibul Bashir, Smruti R Sarangi

ARXIV 2020

Design Space Exploration of Algorithmic Multi-Port Memories in High-Performance Application-Specific Accelerators

Khushal Sethi

RESEARCH ABSTRACTS 2020

Reconfigurable cnn processor for compressed networks

A Ji, W Jung, J Woo, K Sethi, S Luc, A Chandrakasan

ISCAS 2019

Optimized implementation of neuromorphic HATS algorithm on FPGA

Khushal Sethi, Manan Suri

NOCS 2019

Power efficient photonic network-on-chip for a scalable GPU

Janibul Bashir, Khushal Sethi, Smruti R Sarangi

BIOCAS 2018

Low-power hardware-based deep-learning diagnostics support case study

Khushal Sethi, Vivek Parmar, Manan Suri

05 CONNECT

Let's build something
that matters.