Larry Rafsky Email & Phone Number

Jupiter, FL, US

New York, NY, US

I work with a talented team of developers, subject-matter experts, and leading industry consultants, applying deep learning methods to boost productivity in the creation of proprietary financial data and business news products. We specialize in reducing the friction too often encountered in human-machine interaction. We also accelerate the normalization.

New York, NY, US

At Moody's Analytics I initially worked on a variety of AI/ML/NLP algorithms, and I also assisted in the integration of Acquire Media into Moody's.

At ESG-RATE (formerly my4 corp) I direct the team's use of both statistical models and machine learning techniques to mine and analyze the large datasets we use to score public companies on ESG and other criteria, re-weighted and personalized for each user based on the user's expressed social concerns. This includes auto-categorization via topic modeling.

Continuing my Acquire Media work (see two previous entries). Current areas of research include ensemble classification and fact extraction (multi-sensor event and entity detection via data fusion), automated QA methods for NER, and top news/fake news identification via unsupervised machine learning (focusing on real-time big data clustering).

Finding applications for AI, especially ML supported by NLP, to help news publishers and news consumers, focusing on content management (including newswire/photo service/social media/UGC ingestion as well as delivery, distribution, and syndication).

Joined three other founders, and four motivated staffers, to grow Acquire Media from 8 employees to 75. Acquired Newsedge from Thomson Reuters to cement a top-3 industry position in financial news syndication. Developed, in addition to a new version of Newsedge, a real-time news trading application, NewsTrade / v8, and a newsfeed/photofeed content.

Real-time financial news syndication and analysis. Developed XMLNews, a widely-used fusion of NITF for tagged text and RDF for metadata. Integrated news with stock prices to automate a number of trading applications, and built current-awareness models, reports, and graphics for investment banking. As a Dow Jones subsidiary, built the Telerate Interactive.

Directed the design and coding of FAME, a pioneering 4GL for time series analysis. Innovations include true business day, weekly, bi-weekly, semi-monthly, and 4-4-5 data, with exact business-day calculations, holiday calendars, and spline-based interpolation for frequency-shifting conversion of irregular calendar-based data while maintaining correct.

Roseland, New Jersey, US

Time Series and Econometric Forecasting. Founded Econolyst, a joint venture between ADP and Townsend-Greenspan. Developed computing methods to automate and support Alan Greenspan's macroeconomic data collection, forecasting models, graphics, and newsletters, as well as on-line delivery of these products to U.S. corporations.

New York, NY, US

Built quick-turnaround financial and statistical models to support bankers worldwide. Did global tax optimization via nonlinear programming, automation break-even studies via logistic regression, budget-vs-actual and forecast accuracy studies via time series methods, and especially loan-spread profitability studies via LIBOR trajectory simulations.

Applied Statistics Research (Department 17 in Area 10). Focused on Isotonic Regression, Ridge Regression, Multivariate 2-Sample Tests. Also worked with the original CART team.

Under sponsorship from Naval Air Development Center, and Office of Naval Research, did work on ASW (OP95C) tactics via simulation studies, including contributing algorithms to APSUB, APSURF, and APAIR. Became an expert on P3C Orion flight path optimization for reconnaissance work. Modeled SLIR, FLIR, radar and photo results, including PI error studies..

Phd, Statistics

A.B. Summa Cum Laude, Statistics