# Plumerai

- **Website:** <https://plumerai.com>
- **Primary alias:** `plumerai.com`
- **Founded:** 2017
- **Employees:** 11+
- **Location:** London, England, United Kingdom
- **Company kind:** PRIVATELY HELD

## Description

Plumerai is a pioneering brand that specializes in developing software building blocks for embedding production-worthy artificial intelligence (AI) into various products. With a comprehensive focus on the entire AI stack, from data collection to training algorithms, model architectures, inference engines, and hardware optimizations, Plumerai ensures top-notch AI solutions. One of their standout offerings is their people detection AI, renowned for its exceptional accuracy across different environments, all while consuming minimal resources. Notably, their inference engine for microcontrollers has been recognized as the fastest and smallest globally, as confirmed by MLPerf. Plumerai's innovation extends to their Ikva AI accelerator, designed to fit even the smallest FPGA devices. Plumerai's devotion to making deep learning accessible everywhere is evident in their commitment to efficiency, enabling inference on low-power hardware. They have revolutionized the industry by developing tiny technology that operates on $1 chips with less than 10 milliwatts of power consumption. The benefits of processing data locally include enhanced reliability, faster response times, cost-effectiveness, privacy, security, and reduced energy consumption. The Plumerai product lineup features remarkable offerings that leverage their advanced technology. Their people detection AI solution, trained with over 30 million images, provides accurate results, even running on Arm Cortex-M microcontrollers or Arm Cortex-A CPUs with minimal processor load. Additionally, Plumerai's inference engine accelerates any neural network, delivering an average speedup of 2.6x, a RAM reduction of 2.0x, and a code size reduction of 3.6x without compromising accuracy. They also excel in the field of Binarized Neural Networks (BNNs), achieving superior efficiency by using just one bit for each parameter, reducing memory requirements and computational complexity. Plumerai's commitment to technological excellence extends to their ecosystem of open-source Python packages called Larq, specifically designed for BNNs. They optimize their neural networks for small embedded devices and ensure reliability and accuracy through their meticulously collected and labeled datasets. Their software is highly efficient on microcontrollers and seamlessly integrates with ARM Cortex-M, ARM Cortex-A, and RISC-V processors. For customers utilizing FPGAs, Plumerai offers a custom IP-core specifically optimized for running any AI model with maximum energy efficiency. Discover the transformative power of Plumerai today! Their full software solution for people detection and other AI tasks, along with their cutting-edge inference engine and IP-core for FPGAs, is readily available to customers looking to harness the potential of AI in their products.

## Industries

- 🖥 Computer Hardware _(under Computers Electronics and Technology)_
- 🖥 Computers Electronics and Technology

## Links

- [github](https://github.com/larq)
- [linkedin](https://linkedin.com/company/plumerai)

## Logos & icons

- logo _(primary)_ — PNG, SVG — [download](https://cdn.brandfetch.io/idg4R2T4uo/w/820/h/220/theme/dark/logo.png?c=1bxid64Mup7aczewSAYMX&t=1782683324071)
- icon _(primary)_ — JPEG — [download](https://cdn.brandfetch.io/idg4R2T4uo/w/400/h/400/theme/dark/icon.jpeg?c=1bxid64Mup7aczewSAYMX&t=1782683324149)

## Colors

| Name | Hex | Theme |
| --- | --- | --- |
| Dodger Blue | #4080ff | accent |
| Blue Charcoal | #000816 | dark |
| White | #ffffff | light |

## Fonts

- Avenue X — asset _(custom)_
- FK Grotesk — asset _(custom)_

## Imagery

- banner _(primary)_ — JPEG — [download](https://cdn.brandfetch.io/idg4R2T4uo/w/2256/h/382/idBKZzAfxJ.jpeg?c=1bxid64Mup7aczewSAYMX&t=1782683324310)

---

_Source: <https://brandfetch.com/plumerai.com>_
_See [/llms.txt](/llms.txt) for a full list of agent-readable pages, and [/auth.md](/auth.md) for how agents authenticate._