Skip to content
5 min read

Scrypt AI Spins Off from MIT Machine Learning Company

Scrypt AI Spins Off from MIT Machine Learning Company

CheckAlt partners with MIT big data company to strategically deliver cognitive automation solutions, including machine-learning document processing and payments.


Meaningful and relevant information is often buried in image files, scanned documents and PDFs. As a result, significant and valuable data is often “hidden” and inaccessible to automated data analysis. Unfortunately, off-the-shelf optical character recognition (OCR) and text extraction systems are typically inadequate, resulting in solutions that are not scalable or error prone. Although custom solutions can be built using human-generated document templates and up-front annotations, taking this approach can be complicated and time-consuming, often requiring significant capital investment.

The amount of paper document processing across financial services is especially intense and is therefore also an area of tremendous opportunity for efficiency gains and expense savings. Streamlining document sorting, classification and information extraction offers a significant lift in efficiency, and further integration to achieve the “last mile” of an actionable step, whether it is depositing checks, moving cash across accounts, or paying invoices.

MIT CSAIL Alliance member Composable Analytics, Inc. has operationalized an end-to-end AI data capture platform for high-volume, high-accuracy document processing and document discovery. “By partnering with CheckAlt and launching Scrypt AI as a new fintech joint venture, we are taking automatic document digitization to the next level and facilitating business workflows,” said Andy Vidan, Ph.D., CEO of Composable Analytics.

The Challenge

More than half of all invoices received by businesses in the U.S. today are paper, while more than half of business-to-business payments are made electronically. There is a clear relationship between the number of invoices received electronically and the business cost of invoice processing, which is $11.57 per invoice according to 2019 research from Ardent Partners.

Those costs, which include labor, overhead, and technology, are tied to the manual effort of entering invoices into enterprise resource planning systems and then matching each invoice to the correct payment account. At scale, one can imagine how arduous and costly payables management can become.

Though invoicing is becoming less paper-based than before, and electronic payments are accelerating faster due to COVID-19, technology surpassing standard data extraction tools is needed to automate payables, which includes straight-through processing of invoices (either PDF or Word documents attached to email, or physical paper invoices received in the mail) without human intervention.

The Solution

Scrypt is an intelligent invoice coding solution that utilizes machine learning and automation to add and manage invoice payments within your enterprise resource planning software—solving a critical challenge within the payables component of treasury management. Whether businesses receive invoices digitally or in the mail, Scrypt can help them automate payables at a time when they need it most. Scrypt easily integrates with existing ERP systems, such as property management software Yardi’s Voyager, and can automatically identify and match an invoice to the appropriate payment account.

Built on Composable Analytics’ platform with business value input from CheckAlt, Scrypt is the solution for intelligent document processing and information extraction. Scrypt provides an intelligent, accurate and scalable analytics environment for extracting hidden information found in documents and images. Scrypt uses state-of-the-art OCR libraries, image processing routines, and text analytics interchangeably to perform the necessary data extraction with high accuracy and at scale.

The initial applications for Scrypt include invoice coding and payment processing for a variety of industry sectors, including healthcare, utilities, real estate, municipalities, non-profit, higher education, and small business. “Many organizations still struggle with manual business processes, multiple systems, and repetitive, recurring tasks,” said CheckAlt’s Co-Chairman and CEO Shai Stern. “Through our alliance with Composable, our Scrypt platform will further strengthen CheckAlt’s solutions in the payment data capture and integration space. Scrypt technology will be battle-tested and scaled for the benefit of all our clients.”

With Scrypt, businesses can achieve true digital transformation and expense savings by leveraging cognitive automation and intelligent augmentation for high-volume document processing. Furthermore, with CheckAlt’s integrated solutions to streamline billing and payments. Full automatic accounting and reconciliation of accounts are streamlined while reducing tasks and errors.

Scrypt uses a composable machine learning and AI engine to process documents and extract information automatically. No up-front annotations, templating, or configurations are needed, as the system learns from historical data and ongoing extracts. The onboarding process includes training Scrypt on your historical data and, where appropriate, integrating Scrypt with your existing system of records.


At MIT, Lars Fiedler and Andy Vidan, co-founders of Composable Analytics, Inc., met while working on distributed systems and big data projects for homeland security and defense applications. They recognized early on that an army of humans, in different organizations and different roles, all were trying to solve basically the same problems of sifting through floods of incoming information. They decided to build a big data platform that is intelligent, to quickly ingesting the data and analyze it, all in a user friendly, intuitive way.

Recognizing that CheckAlt, an item and payment processing company based in Los Angeles with a nationwide network of processing centers across the U.S., has become the universal payment processing hub for hundreds of financial institutions and businesses, a partnership between the two companies was the ideal approach to scaling out Composable Analytics’ technology.

“We’re truly excited to partner with CheckAlt to deliver an end-to-end, intelligent, high-volume processing system through Scrypt,” said Vidan. “We’re using advanced analytics techniques including data mining, text analytics, natural language processing and computer vision, to digitally sort, process and code documents.”

As both bank agnostic and demographic agnostic, CheckAlt’s mission is to process check payments while also converting paper payments to electronic using robust electronic payment APIs. CheckAlt’s solutions include lockbox processing across a nationwide hub-and-spoke network of processing sites, consolidated item processing, integrated treasury solutions, and mobile capture services.

Composable and CheckAlt launched Scrypt AI to meet the growing demand for payables automation, solving for manual processes around handling, analyzing, importing, paying, and reconciling invoices. “At CheckAlt, we are always looking to stay at the forefront of technology when fulfilling our ongoing client commitment to break new ground together,” said Bobby Rahmanian, Chief Product and Innovation Officer of CheckAlt. “Big data, artificial intelligence, and data-driven decision-making is the future and we are thrilled to partner with Composable on Scrypt to continue to bring innovative solutions to market.”

To learn more about Scrypt, visit