deci Raises $21M to Advance AI Model Development

The Challenge of Deploying AI Models

Successfully implementing AI algorithms necessitates not only sufficient data for training but also appropriate hardware for execution. A frequent disconnect exists between the theoretical potential of data science and its practical application.

Often, the envisioned capabilities surpass what is realistically achievable with available resources. This gap hinders the deployment of AI solutions in real-world scenarios.

Deci: Bridging the Gap Between Theory and Practice

Deci, a startup specializing in deep learning, has developed a platform designed to address this challenge. Their technology focuses on creating models that are compatible with existing data and hardware infrastructure.

The company recently secured $21 million in Series A funding, driven by strong demand from Fortune 500 companies utilizing AI-powered video and computer vision services.

Funding and Investors

Insight Partners led the funding round, with participation from previous investors Square Peg, Emerge, and Jibe Ventures.

New investors include Samsung Next, Vintage Investment Partners, and Fort Ross Ventures. Deci previously raised $9.1 million in a seed round led by Square Peg and Emerge.

Collaboration and Technological Foundation

Deci maintains close relationships with organizations beyond formal investors. Intel collaborated with them on MLPerf, demonstrating the acceleration of the ResNet-50 neural network on Intel CPUs using Deci’s technology.

The platform is built upon Deci’s proprietary AutoNAC (Automated Neural Architecture Construction) technology, enabling rapid model building and continuous updates.

Focus on Computer Vision

Initially, Deci concentrated on computer vision applications, where AutoNAC facilitates the efficient creation of models for services that would traditionally require extensive development and experimentation.

A prominent client, a leading videoconferencing provider (name withheld), leverages Deci to implement AI-powered background blurring during video calls.

This processing occurs “at the edge,” directly on users’ CPU-based devices, which are not typically optimized for AI workloads.

Expanding into Natural Language Processing

Yonatan Geifman, Deci’s CEO and co-founder, announced plans to extend the platform’s capabilities to natural language processing (NLP).

This expansion will support applications such as voice interfaces, personal assistants, audio search, and customer service chatbots.

Addressing Internal Computing Needs

While Deci assists companies in deploying AI on diverse devices, it also caters to organizations seeking to optimize models for their existing infrastructure, even with access to GPUs and substantial computing power.

This reflects a common dynamic in enterprise IT, where businesses strive to maximize the utilization of current assets while simultaneously considering investments in newer technologies.

The Race for Larger Models

“There is a race to larger models all the time,” Geifman stated, referencing Nvidia and Microsoft’s recent language model announcement as an example.

He emphasized that hardware advancements alone cannot keep pace with the increasing complexity of AI models, necessitating a convergence based on available resources.

Deci aims to bridge this gap by optimizing models for specific hardware constraints.

Data Augmentation and Security

Recognizing the ongoing challenge of data scarcity in AI, Deci employs synthetic data generation to supplement existing datasets.

Crucially, all processing occurs within the client’s secure developer environment, ensuring data privacy and preventing external access.

Introducing DeciNets

Deci continues to develop new products leveraging AutoNAC, including DeciNets – a family of computer vision models designed to reduce computational requirements by streamlining the model-building process.

Industry Recognition

Lonne Jaffe, managing director at Insight Partners, praised Deci’s innovative technology and its ability to optimize deep learning models for diverse hardware platforms.

“We are delighted to be part of Deci’s ScaleUp journey and look forward to supporting the company’s rapid growth,” Jaffe stated, and will be joining the board of directors.