During a research internship at SAP Leonardo in 2018, I co-authored a set of patents describing a Deep Learning system for matching financial data to automate corporate accounting processes. Today, this system is powering Cash Application, SAP’s flagship ML product handling over € 200 Million Euro in annual sales pipeline for global companies. I was awared the Professional Internship Book Prize by NTU for my performance and also wrote a more detailed blogpost about my experience at SAP.
Briefly, we developed a pipeline involving: (1) Deep neural networks for building embeddings of structured financial data such as bank statements and invoices; (2) Extracting a semantic similarity graph of financial entities based on the learnt embeddings; and (3) Applying combinatorial optimization techniques for clique finding on the resulting similarity graph.
Sean Saito, Truc Viet Le, Chaitanya K. Joshi, and Raja Shanmugamani
US Patent App. 16208681
tl;dr Representing sets of financial statements and invoices as bags-of-features.
Sean Saito, Chaitanya K. Joshi, Raja Shanmugamani, Truc Viet Le, and Rajesh Vellore Arumugam
US Patent App. (pending)
tl;dr Learning low-dimensional vector spaces of financial statements and invoices for constructing semantic similarity graphs.
Truc Viet Le, Sean Saito, Chaitanya K. Joshi, and Raja Shanmugamani
US Patent App. 16210070
tl;dr Matching financial statements and invoices via graph combinatorial optimization algorithms on semantic similarity graphs.