Google’s Gradient powers Send AI to help businesses extract data from complex documents

The fledgling Dutch startup wants to help companies with additional data from large volumes of complex documents where accuracy and security are paramount — and it has secured the backing of Google’s Gradient Ventures to do just that.

Send AIas the startup is called, it takes on established holders in the document processing space such as UiPath, Abbyy, Rossumand Kofaxwith a customizable platform that allows companies to fine-tune AI models for their own individual data extraction needs.

For example, a company operating in a highly regulated industry such as insurance will likely need to handle a myriad of formats, from PDFs and paper files to smartphone photos taken with all possible orientations and background «noise». Such non-standard «unstructured» data types may be difficult enough for humans to parse, but a fully machine-driven approach can lead to erroneous claim denials or reimbursements and administrative headaches.

Indeed, typical off-the-shelf document processing software is often designed for more common document types that cut across multiple industries, making them unsuitable for certain use cases. With Send AI, on the other hand, companies can train a computer vision model to recognize specific documents and a separate language model to extract and validate relevant data — with humans involved if there’s any doubt to control and review every step through the web interface.

«This validation can be as simple as checking that the expected number is indeed a number or more sophisticated looking up the registration number in a database to see if there is a match,» Send AI founder and CEO Thom Trentelman he told TechCrunch. «Any uncertainty will be reported for human review.»

Recommended Article
Byju's says investors don't have a vote to remove founders from the edtech group

Founded in Amsterdam in 2021 initially as Autopilot, Send AI previously raised a small investment of $100,000 from an alumni fund, but as it started to accelerate things, it has now raised an additional €2.2 million ($2.4 million) in pre – a seed funding round jointly led by Google’s Gradient Ventures and Keen Venture Partners, with participation from a number of angels coming from companies such as DeepMind.

How does it work

Companies can access Send AI’s cloud-based software through APIs that route data from documents sent via email. Upon receipt, Send AI visually enhances documents before sending them to its language models for classification and extraction.

As for the target market, Trentelman says the company is targeting larger businesses, as they «struggle the most with documents,» although really any business that processes large volumes of documents could find applications for the technology

Submit AI: Data Extraction

Image credit Submit AI: Data Extraction

Perhaps it should go without saying that, in addition to a number of existing document processing tools already on the market, Send AI is facing a new breed of startups selling services built on powerful new large language models (LLMs) such as OpenAI with GPT- X (powered by ChatGPT). But while Trentelman admits that such products work great in situations that require a «subjectively good» result such as summarizing or answering questions, where a high degree of accuracy is required in large volumes of documents, that’s a different story.

«You’re going to be up against these technologies soon — big, generic LLMs are still unpredictable, slow and expensive,» Trentelman said. «At Send AI, we let the client build their own solution.»

Recommended Article
Non-sexy industries can also attract investors

Under the hood, Send AI is built on smaller, open-source models that are first trained by the customer by manually processing a small set of documents, then rinsed and repeated on new documents with the help of human proofreaders.

In terms of pricing, Send AI charges on a per-credit basis, with users paying per processing step. «This way we can differentiate between processing a 50-page PDF or just a snippet of a single text,» Trentelman said. «Our models are cheap, fast and reliable, so we can implement them for every customer. This way, clients are in control of their data and performance, which is why we do well in regulated industries like health insurance and government.”


Send AI claims its technology will appeal to highly regulated industries because of the control it gives users over their data, which might seem counterintuitive considering it’s all cloud-based. However, Trentelman points out how a typical LLM like OpenAI works, vis à vis the way it could combine training data from multiple different users into a single model, increasing the potential for leaking sensitive data. Exactly because of that we saw tons of startups dive out with a promise to protect private data within the software powered by LLM.

Send AI tries to solve such problems by setting up small, isolated, open-source transformer models for each user.

«We use a variety of them to get the job done—out of the box they don’t make a big impression, but once trained on high-quality data, they become powerful and accurate,» Trentelman said.

So while the models and associated training data still live in Send AI’s cloud, using isolated models means it can pinpoint exactly where the data is and therefore delete it on demand. That, according to Trentelman, is enough to make it a «preferred candidate» over other providers, and goes some way toward convincing data privacy-minded companies that on-premise deployment isn’t their only option.

Recommended Article
Indian Supreme Court allows Byju to go ahead with crucial EGM

«Today, more regulated companies allow vendors to use the public cloud, as long as they adhere to an extensive list of regulations,» Trentelman said. «We were always asked up front if we could do an on-premise implementation, but in the end all but one company accepted our public cloud offering.»

For now, Send AI is in private beta mode, though it already has some impressive customers, including insurance giant Ax. With a team of seven today, the company plans to use its new cash injection to double its headcount throughout the year ahead of a full commercial launch.

#Googles #Gradient #powers #Send #businesses #extract #data #complex #documents

What do you think?

Deja una respuesta

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *

GIPHY App Key not set. Please check settings

NODWIN Gaming elevates entertainment portfolio with acquisition of Comic Con India

Comfort on the go: The rise of pain relief in the US