We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 - 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
Kyndi, a startup developing what it describes as an explainable natural language processing platform, today announced that it’s secured $20 million in series B funding led by Intel Capital, with participation from UL Ventures, PivotNorth Capital, and existing investors. The capital infusion comes after an $8.5 million series A in August 2017 and brings the company’s total raised to nearly $30 million, and it’ll fuel the expansion of the engineering and sales teams, according to CEO Ryan Welsh.
“As AI becomes an increasingly integral part of how organizations operate and make decisions, there is a growing realization that they can no longer rely on ‘black box’ solutions that lack the rationale behind how recommendations were determined,” said Welsh, who founded Kyndi in 2014. “Some of the leading organizations in government and the private sector are adopting Kyndi’s explainable AI platform because we provide actionable intelligence that’s auditable and provides the reasoning behind every decision.”
Kyndi leverages a battery of natural language understanding techniques to expedite the processing of “massive” amounts of data, all while ensuring its analyses remain auditable and explainable such that the results are justifiable. The company’s tech provides contextual answers to natural language questions and extracts signals from text, delivering insight and surfacing knowledge in previously abstruse corpora.
Kyndi’s secret sauce is a decades-old programming language called Prolog that was designed with probabilistic and logical computational reasoning in mind. Using very little training data — in some cases as few as 10 to 30 documents of 10 to 50 pages each — Kyndi can automate fact, inference, and concept generation in virtually any vertical. It’s speedy to boot (the company claims it can read upwards of a thousand documents in seven hours) and helps to mitigate bias that might otherwise arise from conventional deep learning approaches.
The system works by tokenizing text and identifying parts of speech and sentence structure before identifying named entities with real-world references, and then computing semantic distances among words using a proprietary model. The entity relationships are subsequently parsed and named, creating a so-called proto-ontology that encodes the meaning of documents, repositories, and domains in a queriable knowledge graph.
Kyndi counts clients in government segments including defense intelligence, as well as finance, health care, IT, and infrastructure, and it claims its trainable models are being used in conjunction with robotic process automation tools from UiPath, Automation Anywhere, and BluePrism. Welsh says its tools help to evaluate data integrity and quality issues, detect fraud, and review audits..
“Enterprises are turning to AI to take advantage of new opportunities and to solve pressing business problems, and we expect AI’s use in business will continue to grow as the technology matures,” said Intel Capital senior managing director Nick Washburn. “Solutions like Kyndi’s, which remove some of the mystery of AI technology, will continue to gain importance, and we look forward to helping them accelerate AI adoption and address the need for explainability.”
Kyndi’s other backers include Citrix’s Startup Accelerator and Darling Ventures.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.