Tax Law. Software Engineering. Artificial Intelligence.
Since 2013, we have focused on helping large CPA practices solve complex issues relating to transaction analysis and partnership tax compliance.
We build intuitive, human-centric solutions that elevate quality, increase productivity, and reduce delivery costs.
Our team unites experienced software engineers, accounting and finance professionals, and cloud-computing specialists.
We take pride in our deep business knowledge, empathy for customers, openness to new ideas, and commitments to active listening and critical thinking.
In 2023, systems invented and built by Crimson Tree Software service over $2 billion in recurring annual fees.
Development of a new and practical artificial intelligence system for comprehension of K-1s, K-1 footnotes, and other associated tax documents.
Integrating multiple strategies including generative AI, Bayesian inference, pattern recognition, self-programming, heuristic algorithms, and traditional software automation, this system understands and extracts data from complex documents with unprecedented speed and accuracy.
In the simplest use, users can pass in a single PDF document, or set of documents, and receive a data package with annotated PDF pages and Excel files containing all extracted data organized in regularly formatted tabs.
In interactive use, users can search, sort, filter, slice, and dice data across their document repository, and can query their vectorized text through a ChatGPT interface.
Our Tax AI can comprehend hundreds of semantic concepts and data types, including but not limited to:
The sale of this system to an international accounting firm was completed in 2022 and by 2023 yielded a nine-figure return on investment.
This system comprehensively addresses the needs of a large partnership tax practice, including the following key features:
Among other projects, in 2019 the system was used to propose and model the Up-C conversion of a prominent publicly traded entity. Upon conversion, the entity enjoyed a significant increase in market value. Other businesses with similar structures followed.
Core functions include:
These projects succeeded before the advent of modern Large Language Models and Generative AI.
A difficult area in which we made significant progress was the classification and extraction of data from complex footnotes that trigger reporting requirements (e.g., forms 926, 8865, 8886, and 8621).
By 2020, these systems were trained to recognize and extract data from 44 distinct classes of footnote.
The core libraries united open-source software, closed-source commercial software, and custom developed software to produce a unified and royalty-free system for extraction of data from tax documents.
Examples of deployed applications include:
Clients include prominent hedge funds and regulated investment companies. Many individual client portfolios have more than 1 million transactions subject to analysis each tax year. Read More
The core tax adjustment calculations include:
This work yielded a new, high-performance system of algorithms to compute US partnership tax allocations pursuant to IRC Subchapter K.
In 2013, a Big 4 accounting firm purchased this partnership allocation engine and portfolio of related inventions.
Our team was then engaged in a successful multi-year project that resulted in a comprehensive, “GAAP to Tax” partnership tax compliance system.
Our contributions include:
Today, our systems remain at the core of the firm’s partnership tax compliance practice, with over 6,000 daily users during peak periods.
Maple Lawn, MD