Nomadic Raises $8.4 Million to Structure Data From Autonomous Systems

Companies building autonomous systems continue to generate large volumes of video data. These include self-driving cars, industrial robots, and construction equipment. However, most of this data remains underutilised due to the complexity of processing it at scale.

Traditionally, teams rely on manual review to organise and label footage. Even with shortcuts like fast-forwarding, the process remains slow and resource-intensive. As a result, a significant portion of fleet data often stays archived and unused.

Focusing on rare but critical scenarios

The challenge becomes more complex when companies try to identify edge cases. These are rare events that play a key role in training machine learning models.

For instance, scenarios such as unusual traffic behaviour or unexpected environmental conditions are difficult to locate. Yet, they are essential for improving the reliability of autonomous systems. Without efficient tools, extracting such data becomes a bottleneck.

Nomadic’s approach to structured datasets

NomadicML is building a platform to address this issue. The system converts raw video footage into structured and searchable datasets.

It uses a combination of vision-language models to analyse and categorise events within videos. As a result, companies can search for specific scenarios instead of manually reviewing hours of footage.

This approach supports better fleet monitoring. In addition, it helps teams create targeted datasets for reinforcement learning. Consequently, development cycles can become faster and more efficient.

Funding to expand platform capabilities

The company has raised $8.4 million in a seed funding round. The round was led by TQ Ventures, with participation from Pear VC and Jeff Dean.

With this funding, Nomadic plans to onboard more customers and further develop its platform. Recently, the company also won first prize at Nvidia GTC’s pitch competition.

Founders’ background and product vision

Nomadic was founded by Mustafa Bal and Varun Krishnan. The two met while studying computer science at Harvard.

Before starting the company, both founders worked at firms such as Lyft and Snowflake. During this time, they encountered similar challenges related to processing large-scale operational data.

The platform is designed to help organisations extract insights from their own datasets. Instead of relying on generic data, companies can train models using scenarios specific to their operations.

Use cases across autonomous systems

The platform enables teams to identify specific situations within video data.

For example, it can detect instances where a vehicle proceeds through a red light under police direction. Similarly, it can isolate patterns such as vehicles passing under certain types of infrastructure.

These insights can support both compliance requirements and model training workflows.

Early adoption by industry players

Several companies are already using the platform. These include Zoox, Mitsubishi Electric, Natix Network, and Zendar.

According to industry feedback, the tool helps reduce reliance on outsourced data annotation. It also enables faster scaling of machine learning workflows.

Competition and emerging trends

The market for data annotation and processing tools is evolving. Companies such as Scale, Kognic, and Encord are developing AI-based solutions.

At the same time, Nvidia has introduced open-source models designed for similar use cases.

Nomadic, however, positions its platform as more than a labeling tool. The system aims to function as a reasoning engine that can interpret and contextualise actions within data.

Looking ahead: beyond video data

The company is now working on expanding its capabilities. Future developments may include tools for processing non-visual data such as lidar sensor inputs.

In addition, integrating multiple data sources remains a key focus area. This could improve how autonomous systems interpret complex environments.

Handling large-scale datasets continues to be a technical challenge. Processing vast volumes of video alongside advanced machine learning models requires significant computational resources.

However, as autonomous systems evolve, tools that can structure and analyse data efficiently are likely to play an important role.