Keshav Pingali, CEO and co-founder of Katana Graph, was interviewed by JP Valentine on the March 30th AI in Action podcast, which explores the impact that Data Science, Machine Learning, and Artificial Intelligence are making on our everyday lives.
To introduce Katana Graph, Keshav explained that about a decade ago some colleagues got together and began building the components for high-performance graph computing. Their first project, coordinated by DARPA (Defense Advanced Research Projects Agency), was designing a real-time intrusion detection system in computer networks using provenance graphs to identify and flag forbidden activity for BAE Systems. Katana Graph took shape in April 2020 and since then has grown to about 100 employees, all striving to solve two critical challenges facing modern enterprises: the volume of data generated and the time to insight.
The tremendous amount of data accumulated across the world is mostly unstructured but can be usefully represented as graphs containing entities and the relationships between them — nodes and edges. With the pressing need across industries to quickly process and analyze massive volumes of data to gain insight within a window of opportunity, Katana Graph lets users quickly and easily extract knowledge from very large amounts of unstructured data.
In the pharmaceutical industry, Katana Graph is bridging the gap between medical knowledge and the tools available to data scientists. The Katana Graph intelligence platform can aid in tasks such as hypothesis generation, which aims to generate or rule out potential treatments for a given disease by inspecting a medical knowledge graph before spending time and resources in a lab. Some key capabilities of Katana Graph’s graph engine are its exceptional rate of data ingestion, which lends itself well to any application that requires a great deal of data, and its AI-enhanced analytics to get accurate results out of data.
Data volume and time to insight are critical in applications such as preventing unauthorized activity and identifying fraud rings. Graph technology in the banking sector has made big strides in reliability and security, but many systems in this area employ “old-fashioned” AI and are typically rule-based systems that have limited ability to see beyond their constructs. Building a graph representing everything that is known about past financial transactions reveals relationships and tendencies among the data, and when a new transaction is presented to the machine, a prediction can be made regarding its authenticity. With the already vast amount of past transactional data constantly growing, the need for a scalable solution is urgent.
Financial institutions and related verticals use Katana Graph’s graph platform to perform queries, AI, and analytics on data at an unrivaled scale, recently tested up to 256 machines, the maximum available in the test setting. By integrating mining, querying, analytics, and graph neural networks (GNNs), this comprehensive graph intelligence platform evaluates the rich information provided by relationships at an unprecedented rate. The inclusion of graph AI enables GNN training, which in turn allows the platform to make timely inferences and essential predictions.
Katana Graph has a strong focus on ease of use and has created a graph data platform that can interoperate with third-party libraries and be integrated as cloud-native or installed on premises. In essence, to run an application, the user indicates how many machines are available for use and the platform partitions the given graph into prescribed or custom shards that fit in the memory of each machine. Local in-memory graph computing engines provide a library of data structures and runtimes to perform computations on the worker machines while a communication runtime ensures synchronization among partitions. The features of the platform are exposed to allow users to orchestrate the graph engine using familiar languages such as Python, C++, Cypher, and GQL.
Keshav concluded by saying that Katana Graph is eager to bring their technology into the Healthcare and Fintech spaces and continue to work closely with clients to solve problems in the ever-expanding world of tech.
