SiMa.ai, a leader in Physical AI, has launched Palette
Neat™, described as the industry’s first agentic development environment for
Physical AI, designed to reduce complex application timelines from months to
days.
The open-source, purpose-built integrated development
environment combines a Physical AI execution library with an agent workflow
layer for productivity-focused agentic development. When paired with the
full-production Modalix™ MLSoC™ System-on-Module (SoM), or its new PCIe
companion card form factor, the unified platform delivers performance-per-watt
for high-demand Physical AI workloads across robotics, automotive, drones,
industrial automation, aerospace and defence, smart vision and healthcare.
“SiMa.ai is an AI software company that builds its own
silicon,” said Krishna Rangasayee,
Founder and CEO of SiMa.ai. “Today, we are delivering the industry’s first
agentic development environment for Physical AI. Together, Palette Neat and our
pin-compatible SoM dismantle the incumbent GPU moat, allowing developers to
design systems in plain English and develop them in days — and in many cases,
hours.”
Palette Neat uses a natural-language interface and an
agentic workflow to abstract low-level compute complexity. This reduces the
months of work traditionally required to port and integrate applications to new
silicon.
Key advantages of Palette Neat include a new development
paradigm: Developers can use natural-language commands to build entire systems,
enabling engineering teams to focus on system-level differentiation for both
new and legacy applications.
From months to days or hours: The agentic environment
autonomously builds and maps applications directly to silicon, shrinking
development cycles significantly. Developers can also reuse existing
application code, preserving approximately 90 per cent of their legacy software
investment.
Frictionless platform migration: Palette Neat, along with
the pin-compatible SoM and new PCIe companion card, is designed to dismantle
the incumbent GPU moat while reducing the cost, time and engineering risk
involved in switching hardware platforms.
The full-production Modalix SoM runs multiple Large Language
Models (LLMs) concurrently alongside vision and sensor models, all under 10W.
Built from the silicon up for Physical AI deployment, it is designed as a
pin-compatible drop-in replacement for the incumbent NVIDIA SoM form factor and
requires no carrier board redesign.
Together, Palette Neat and Modalix SoM aim to reduce the
engineering friction involved in adopting new AI hardware, without requiring
developers to integrate a completely new architecture or rewrite their entire
software stack.
Palette Neat is available as an open-source platform on
GitHub. Documentation is available through the Developer Centre, while the full
hardware specification for the Modalix MLSoC SoM is also available. SiMa.ai
will also host a webinar on 30 June titled “Scaling Physical AI”. Founder and
CEO Krishna Rangasayee has further shared insights in his blog, “Dismantling
the GPU Moat to Scale Physical AI”.