About
The intelligence layer for engineering.
We’re building an AI-native engineering platform that fundamentally changes how complex physical systems are designed — and over time, evolves into a new class of foundation model for engineering itself.
The Shift
From AI as autocomplete to AI as an engineering partner.
An engineer describes a system in natural language — a propulsion assembly, a structural bracket under load, a thermal management loop. The platform translates that intent into structured engineering artifacts, then orchestrates the workflow: generate geometry, apply constraints, run simulations, evaluate performance, and iterate.
Engineering is fundamentally reasoning under constraints — balancing strength against weight, performance against cost, efficiency against manufacturability. Most AI systems operate at the surface level of engineering. We aim to model the underlying structure and physics.
Rather than replacing existing engineering software, CogniCAD integrates with CAD, simulation, and analysis workflows as an intelligence layer that understands both engineering language and physical behavior.
How we’re different
From passive tools to active collaborators
Traditional engineering tools execute commands but do not understand intent. CogniCAD acts as a domain-aware orchestration layer that converts natural-language objectives into structured engineering workflows across CAD, simulation, optimization, and validation systems.
Physics-aware latent space
Most generative design systems produce geometrically plausible outputs without true physical understanding. We represent geometry, constraints, materials, simulation states, and governing equations within a unified latent representation — enabling the system to reason about engineering behavior, not merely generate shapes.
Orchestrator of specialized agents
Complex engineering tasks are decomposed into coordinated sub-problems handled by specialized agents for geometry generation, simulation, optimization, verification, and manufacturability analysis. A central orchestrator maintains cross-stage context and iteratively converges toward designs that are both physically valid and operationally useful.
A horizontal intelligence layer
MMechanical, aerospace, electronics, robotics, thermal systems, and beyond — our goal is to build a shared intelligence layer across engineering disciplines. If frontier AI models are becoming the productivity layer for knowledge work, CogniCAD aims to become the intelligence layer for engineering workflows — accelerating design cycles from weeks to hours.
Founders
Dhruv Chaturvedi
Co-founder
Researching on ai native cad archtectural black boxes like cad tokenizer, multimodal cad transformers and physics-aware latent space at the core of CogniCAD. Background in Aerospace Engineering and applied ML for engineering systems.
Shresth Keshari
Co-founder
Co-founding CogniCAD with a focus on developing the physics-aware latent space, manufacturing and synthesis capabilities. Background in Computational Mechanics, Neural Networks and Machine Learning for engineering applications.
Long-Term Vision
The Large Spatial Engineering Model.
Frontier models excel at language and code, but lack a deep understanding of space. Engineering is, at its core, a spatial discipline — wings, chips, heat exchangers, robots — geometry, topology, and constraints interacting in three dimensions.
LSEM unifies geometry, physics, and constraints in a single latent space. It enables reasoning across structure and physics simultaneously — not just generating designs, but iteratively refining them while explaining its reasoning. Engineering cognition at scale.
Mission
In the near term, we accelerate engineers. In the long term, we redefine engineering itself — the intelligence layer for the machines, systems, and infrastructure that define the real world.