Snorkel AI, originated at the Stanford AI Lab, is a technology startup based in Palo Alto, California. The organization makes use of programmatic data labeling to accelerate the enterprise AI application development and its deployment. Snorkel AI manufactures tools that efficiently manage the data without any hand-labeling, train the models, and analyze and iterate the AI systems. In the recent Series B funding rounds, the company raised a whopping $35 million, taking the company’s total funding to around $50 million. The money raised will be utilized to scale up the engineering team further and attract new customers and clients.
The platform launched by the Stanford AI Lab in the year 2019 has the primary aim of redefining how people look at artificial intelligence. It claims to make it more practical and usable by simplifying one of the most tedious aspects of machine learning: data labeling (required to teach and train the machine learning algorithm) because it proves to be highly costly and labor-intensive. The company says that it has managed to develop software that programmatically adds the labels required and minimizes the subject experts’ and data scientists’ effort.
Snorkel AI has claimed to offer a radically new approach of looking at AI in contrast to the conventional ones being used. The traditional AI approaches rely on generic third-party models wherein several human labelers must feed into the systems. The new approach offered by Snorkel AI enables the users to programmatically label the training data available in mere minutes (can vary based on the complexity of the data).
Snorkel AI works in four straightforward and easy steps:
The large datasets are programmatically labeled instead of hand labeling
All the data is automatically cleaned, combined, and managed by the algorithm
Large and state-of-the-art models are trained with just a click of a button
The platform also allows for analysis of the data and identifying and rectifying any error found
The company claims to offer the following advantages to all of its users:
- Faster development and deployment of AI
- High accuracy
- Collaborative workflow
- Easy to change applications
- Privacy safe labeling
The company recently launched another solution named Application Studio. It is a visual builder with templated solutions for common AI use based on the best practices from academic institutions. This Application studio is expected to give the company an impetus, and it has claimed to introduce prebuilt solution templates for industry-specific use. It will also provide app-specific pre-processors, programmatic labeling templates, and high-performance open-source models, which would be allowed to train with private data. Application Studios is, for now, in a preview mode wherein it will be looked through in great detail.
The data labeling toolset is trained time and again by the company to ensure no data breach and data bias. According to the organization, simplifying the labeling process simplifies the process of detecting biases simultaneously. The system is vigilant and proactive in identifying any of the biases present. The company maintains the data’s privacy in the highest standard possible as it claims that the majority of the information is not even looked at by humans.