Sparse vs. Dense Data
Comparison of Supervised Learning Algorithms: k-NN, SVM, Decision Trees, Boosted Trees, and Neural Networks
Expert witness discovery and recommendation system
Deep learning, web scraping, data pipelines
Beche Group provides expert witness search services to law firms to support patent litigation.
Most law firms find an expert witness by referral and word-of-mouth reputation. A few companies offer digital listings of expert witnesses. Both of these traditional methods don't leverage the data available.
I designed and built a prototype recommendation engine to find expert witnesses. Given a patent, a user receives a ranked list of experts based on case history, patented inventions, and published papers.
I built the data pipelines to support the application. Data is acquired via public APIs and scraping publicly available case histories. This includes patent applications, depositions, curriculum vitaes, and other documents, primarily in PDF form.
Initial tests have benefitted Beche Group. The application identifies previously unknown expert witnesses who are capable of addressing a given patent, moving beyond the traditional methods mentioned above.
Since 2020 I've been the sole individual contiributor on this project. The work described above is the intellectual property of Beche Group, therefore I cannot share the source code.
Other Work
Reinforcement Learning in the Stock Market
Back-testing learned strategies vs. manual strategies in the stock market.