What ModelCat AI does

ModelCat AI is a Sunnyvale, California-based SaaS company that operates an AI-in-the-Loop platform designed to autonomously generate, optimize, and validate machine learning models for edge, embedded, and IoT hardware. Formerly known as Eta Compute, the company rebranded in September 2025 to reflect an expanded scope beyond edge AI to all devices where intelligence must run locally. Its core technology uses a proprietary AI trained on the model development process itself to explore a design space of more than 50,000 neural network architectures, generating thousands of candidate models, testing them against user-defined latency, memory, and power constraints, and iteratively refining outputs until deployment targets are met.

The platform delivers a single-step workflow covering data quality checking, architecture selection, training, optimization, and hardware validation, with typical turnaround of 24-48 hours and contractually positioned delivery in three days or less; customers receive ready-to-run models in TFLite or other standard formats. Product offerings include the core ModelCat Platform, a partner-branded eIQ Model Creator built specifically for NXP Semiconductors, an AI Model Retargetor capability launched in March 2026 for cross-silicon model portability, and a dedicated Hardware Farm that validates models against physical chips from NXP, Silicon Labs, Alif Semiconductor, ST Microelectronics, Qualcomm, and Synaptics. The AI-in-the-Loop methodology is a registered patent, and the Model Lake of prior build outcomes gives the system an informed starting point for new jobs.

Revenue is generated through three streams: subscription access to the AI Model Building Platform at app.modelcat.ai, OEM and embedded licensing through semiconductor partners (anchored by the NXP relationship), and professional services and pilot engagements for enterprise customers. Pricing is quote-based and not publicly disclosed, with deal sizes positioned in the six- to seven-figure range. The company sells primarily to enterprise embedded and IoT product teams, semiconductor vendors, and companies in automotive, industrial automation, and robotics, with headquarters at 182 S. Murphy Ave, Sunnyvale. Leadership comprises CEO Evan Petridis, CTO Jeremi Wojcicki, PhD, and CRO Jon Gettinger; headcount sits between 11 and 50 and the company is venture-backed with no public listing.

Name

ModelCat AI

Legal name

ModelCat, Inc.

Website

https://modelcat.ai

Company type

Private

Founded year

2015

Operating status

Operating

Headcount range

1–10 employees

Short description

ModelCat AI provides an AI-in-the-Loop SaaS platform that autonomously builds and validates custom ML models for edge, embedded, and IoT hardware. It serves enterprise product teams and semiconductor ecosystems, with flagship integration into NXP's eIQ AI toolchain.

Ownership category

akta.pro rank

Headquarters

HQ city

Sunnyvale

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Marketing or Sales, Operations

Revenue model

1. AI Model Building Platform Subscription: ModelCat offers access to its AI-in-the-Loop model building platform as a software service. Users access the platform via app.modelcat.ai to upload data, select target hardware, set constraints, and receive optimized models. The platform includes managed compute infrastructure for training, optimization, and validation, removing the need for customers to maintain dedicated training clusters or specialized infrastructure. Pricing is quote-based and not publicly disclosed.
2. OEM/Embedded Licensing (eIQ Model Creator): ModelCat provides an exclusive version of its platform (eIQ Model Creator powered by ModelCat) developed in collaboration with NXP. NXP customers can access a trial version via NXP's official development kits, with full versions available directly from ModelCat. This represents an OEM/embedded licensing revenue stream where NXP bundles or resells access to ModelCat's capabilities.
3. Professional Services / Pilot Engagements: ModelCat engages in pilot projects with customers, including application engineering reviews of target hardware and use cases, structured test drives using customer data and target hardware, and onboarding engagements. These are delivered alongside the self-serve platform for enterprise customers.

Go-to-market motion3 records

Distribution channels4 records

Marketing channels6 records

Core offering

ModelCat operates an AI-in-the-Loop SaaS platform that autonomously builds, optimizes, and validates custom machine learning models for edge, embedded, and IoT devices. Users upload labeled data, select target hardware, set constraints (latency, memory, power), and receive production-ready models validated on real silicon within 24-48 hours to 3 days. The platform is distributed both directly and via OEM partnerships (e.g., NXP's eIQ Model Creator).

Product overview

ModelCat AI offers a unified AI-in-the-Loop platform called the ModelCat Platform that autonomously builds, optimizes, and validates custom ML models for edge, embedded, and IoT devices. The core platform delivers production-ready models in a single automated step, contrasting with traditional 12–24 month DIY workflows. It integrates the Hardware Farm for real-hardware validation and the AI Model Retargetor capability for cross-silicon model portability. ModelCat also distributes a co-branded sub-product, eIQ Model Creator powered by ModelCat, developed specifically for NXP Semiconductors as an exclusive version of the platform tailored to NXP silicon. The portfolio centers on the principle that constraints (latency, memory, power) are first-class inputs, not afterthoughts.

Differentiator

Problem solved

Functional benefit

Products and services

• ModelCat Platform AI-in-the-Loop autonomous platform that builds, optimizes, and validates custom ML models for edge, embedded, and IoT hardware. Users upload labeled data, pick target hardware, set constraints, and receive ready-to-run models in TFLite or other formats within 3 days or less.
• eIQ Model Creator powered by ModelCat Co-branded version of the ModelCat platform developed in collaboration with NXP Semiconductors and integrated into NXP's eIQ AI Software Development Environment. Delivers custom, ready-to-run ML models for NXP's full silicon portfolio, reducing development cycles from 12-24 months to days.
• AI Model Retargetor Agentic capability that adapts trained AI models to new target hardware without rebuilding from scratch. Iteratively modifies architecture configurations, memory usage, compute scheduling, and optimization strategies to fit models within latency, memory, and power constraints of new silicon.

Quantifiable outcome

• 120ms inference latency achieved vs. 500ms target (4.2x improvement) — Vidar Systems, delivered in 3 days
• +3 more outcomes

Named customers3 records

Segments5 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability9 records

Feature8 records

Partnerships

Six partnerships are on record, tiered core and flagship.

• Silicon LabscoreTechnology or Integration · 5 March 2026Silicon Labs is a supported silicon vendor in ModelCat's platform. The AI Model Retargetor supports EFR32MG24 and EFR32MG26 chips from Silicon Labs. ModelCat is trusted by Silicon Labs as a platform for their customers' edge AI model development. Silicon Labs' EFR32 series chips are among the supported targets listed on the platform.
• Alif SemiconductorcoreTechnology or Integration · 5 March 2026Alif Semiconductor is listed as a supported silicon vendor in ModelCat's AI Model Retargetor platform. Alif's chips are among the target hardware platforms that ModelCat can build and port models for, enabling AI deployment on Alif's edge AI semiconductor products.
• ST MicroelectronicscoreTechnology or Integration · 5 March 2026ST Microelectronics is a supported silicon vendor in ModelCat's platform. The AI Model Retargetor supports multiple STM32 chips including STM32F401RE, STM32F429ZI, STM32F746ZGT6, STM32F767ZI, STM32H723ZG, STM32H7A3ZI, STM32L476RG, STM32L4R5ZI, STM32L552ZE, and STM32U575ZI. ST Micro is listed as a trusted chip partner on ModelCat's demo page.
• NXP SemiconductorsflagshipStrategic or Co-development Partner · 8 September 2025ModelCat launched eIQ Model Creator powered by ModelCat, an exclusive version of its AI model building software developed in collaboration with NXP Semiconductors. Optimized for NXP's i.MX applications processors, MCX microcontrollers, and i.MX RT crossover microcontroller devices. The software expands NXP's eIQ AI Software Development Environment, adding automated AI model creation capabilities. NXP customers can access trial versions via NXP's official development kits, with full versions available from ModelCat. The platform spans NXP's full portfolio, enabling customers to develop AI once and deploy across all NXP devices without changing tools.
• QualcommcoreTechnology or IntegrationQualcomm QBR5165 is listed among ModelCat's supported chips on the platform. ModelCat can build and optimize AI models for Qualcomm's edge AI silicon, enabling deployment on Qualcomm-powered devices.
• SynapticscoreTechnology or IntegrationSynaptics KA10000 is listed among ModelCat's supported chips on the platform. ModelCat supports AI model building and optimization for Synaptics' edge AI semiconductor products.

Scale indicators6 records

Recent moves6 records

Expansion highlights6 records

Direct peers

• Edge Impulse: Edge Impulse is the closest direct competitor — a developer-focused edge ML platform that ingests data, trains models, and validates them for constrained devices. It serves overlapping IoT/embedded developer buyers and competes on the same autoML-for-edge workflow.
• SensiML: SensiML provides an end-to-end AutoML toolkit for IoT edge devices, with a particular focus on time-series sensor data. It serves a similar developer/embedded-systems buyer and competes on automated model generation for constrained hardware.

Broad incumbents

• STMicroelectronics (STM32Cube.AI): STM32Cube.AI is the chip-vendor-native model optimization tool bundled with ST Micro silicon. While broader in scope (hardware plus software), it overlaps with ModelCat on neural-network optimization for ST microcontrollers and is a key competitive alternative ModelCat must displace inside the ST ecosystem.
• Qualcomm AI Hub: Qualcomm AI Hub is the chip-vendor's model optimization and deployment platform for Qualcomm silicon, including the QBR5165 supported by ModelCat. It competes head-to-head for AI workloads running on Qualcomm edge devices and benefits from being bundled with hardware.
• Google Vertex AI (with Edge / AutoML): Google's Vertex AI and AutoML offerings cover model training, optimization, and increasingly edge deployment workflows. As a hyperscaler incumbent, Google competes on AutoML breadth and cloud integration rather than on hardware-aware validation.
• Renesas e-AI: Renesas e-AI is the chip vendor's native AI toolchain for translating and deploying neural networks on its microcontrollers. It overlaps with ModelCat on the model-to-constrained-device workflow and is bundled with Renesas silicon.
• OctoAI (OctoML): OctoAI (formerly OctoML) provides automated model optimization and serving, originally focused on cloud but expanding toward edge. It shares ModelCat's automated optimization thesis but targets a broader ML-ops buyer rather than embedded specialists.

Emerging players

• Deeplite: Deeplite focuses on deep-learning model optimization for edge devices using automated compression, quantization, and architecture search. It overlaps with ModelCat's optimization layer and competes for embedded AI teams that want lighter-weight models.
• Neurala: Neurala builds AI for edge devices including vision models for industrial and robotics use cases. It overlaps with ModelCat on edge-vision AI for embedded customers and competes for similar industrial/IoT deployments.
• Syntiant: Syntiant develops ultra-low-power edge AI processors and tightly coupled neural network deployment tools. While primarily a silicon vendor, its model development environment competes with ModelCat for low-power always-on AI workloads.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat5 records

Key risks6 records

Key highlights6 records

Customer concentration

• Xx.com/ModelCat_ai
• Facebookfacebook.com/ModelCat.OnDeviceAI

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about ModelCat AI