From High-Frequency Trading to AI Innovation
Accrete’s mission is to end organizational knowledge loss by compressing tacit human domain knowledge into knowledge engines that enable AI agents to make superhuman predictions and decisions in perpetuity. Prior to launching Accrete in 2017, Prashant spent over a decade building high frequency trading systems that used rules based algorithms on structured data to arbitrage mispriced securities between exchanges. Prashant believed that the digital explosion would make it impossible for humans to make sense of contradictory narratives, viral influence, and misinformation. Prashant hypothesized that if he could get a computer to read, understand, reason, and learn more efficiently than human experts, he could mine an ocean of predictive alpha buried in unstructured data and scale.
Breakthroughs in Information Compression
The advent of transformers in 2017, a type of deep learning neural network that enables AI models to compress information by creating mathematical representations of context, enabled AI models to learn by example from unlabeled data in an unsupervised manner. Prashant and his team began work to modify transformer models to compress tacit human domain expertise into AI software that could reason at scale, learn from human feedback while online, build upon accumulated knowledge, and extract predictive insight from dynamic unstructured data that human experts couldn’t.
A Billion Epiphanies to Creating Enterprise Value Using Knowledge Engines to Combat Knowledge Loss
To mine unstructured data for alpha, Prashant and his team leveraged deep learning and reinforcement learning breakthroughs to develop AI models capable of reasoning and predicting better than human experts. These models demonstrated their efficacy in extracting alpha in M&A rumors and predicting the virality of commercially viable musical artists from pre-release social chatter. Realizing that Accretes AI knowledge engines could extract alpha in any domain, Prashant and his team developed a platform for scalable development of Knowledge Engines, the engines transform human expertise and data into autonomously constructed knowledge graphs. These self perpetuating graphs are semantic models that produce predictive insight from dynamic unstructured data with superhuman accuracy. Knowledge engines enable AI agents to reason using long term memory, collaborate with other agents in multi agent frameworks, learn continuously through natural interaction, and ultimately make better decisions than human experts.
Trusted by Government, Valued by Business
Today, Accrete is a dual use enterprise AI company that licenses highly specialized high-performance expert AI agents to both government and enterprise customers struggling with information overload and knowledge loss. Accrete’s Knowledge Engines compress tacit domain expertise into AI software that reasons at superhuman scale and generates predictive insight accurate enough to drive automated decisions. Accrete’s customers include the U.S. Department of Defense, the U.S. Special Operations Command, the U.S. Army, and Fortune 500 companies in industries including Media and Entertainment and Consumer Goods.