A Hands-On Guide for Developing and Applying Custom ESG Metrics. The chapter presents several increasingly effective applications of NLP (including entity-recognition, topic-detection, and sentiment) in ESG, as well as novel techniques to identify cases of 'greenwashing.'

Sentiment alone provides an incomplete picture of what's actually happening in corporate Conference Calls. Is management exaggerating? Are they obfuscating? Are the analysts on the call even buying it? We outline a model to 'read between the lines' of management BS.

Creating company peer-groups using topic modelling / clustering on text from quarterly earnings calls and company business descriptions.

Lightweight implementation of Brinson Attribution: understand and decompose return drivers of your portfolio into stock-selection vs. allocation, along any single or hierarchical category.

Contributing author to a collection of Artificial Intelligence (AI) case-studies in Investment Management.

Talk @ CFA Society of Columbus: Hands-on instruction on data sourcing, sentiment models (FinBert vs L&M Dictionary vs Naive Bayes), and contextualized performance analytics.

ETF holdings provide a surprisingly rich variety of base-line ingredients for most any quant-project: a universe of names (across any region or market segment), symbology (name/ticker/cusip/sedol), sector and style assignments (for basic risk control), and benchmarks for active strategies.

Provides 6 years of Drucker ratings across ~800 domestic large-cap companies. The Drucker Institute brings together five dimensions of corporate performance: [1] Customer Satisfaction; [2] Employee Engagement and Development; [3] Innovation; [4] Social Responsibility' [5] Financial Strength.

Scrape and structure the full history of Executive Compensation from DEF 14-A Filings in the of SEC Edgar Database. Due to changing formats and standards (especially before 2010), we apply a Random Forest Classifier to locate the compensation table amongst many other tables and exhibits contained in the same document.

Collecting television show scripts (and ratings) as source data for random machine-learning and NLP projects. e.g. creating/comparing various word-embeddings across different genres.

Systematically identify trending alternative-data providers as well as broader shifts in the vendor landscape.

Entity mapping is especially challenging when dealing with unstructured data. In this example, we show how to leverage Google's Programmable Search Engine to map names in the Russell 1000 to their Glassdoor URL's.

Collecting news articles for all the companies in the R1000, for a pre-defined set of news outlets, using Diffbot's Knowledge Graph.

Implementation and application of OpenFIGI api

Download 20 years of historic (daily) prices from Alpha Vantage and calculate total returns (split-adjusted, w/ dividends reinvested) across varying frequencies (e.g. monthly). Provides optional AWS integration (S3+Athena)

Connecting portfolio performance drivers to high-quality, relevant market commentary can be time-consuming and costly. Off-Script Systems collaborated with Spain Consulting, LLC to design an intelligent, automated financial research tool.

Discerning Portfolio Managers are keen to objectively measure and understand performance drivers. Off-Script Systems joined forces with Heximer Investment Management to design and implement an institutional quality, fully customized attribution platform.

Jointly developed with Cypress Capital Group: [1] a robust, multi-method web-scraping application that cleans and structures property information into analytics-ready data stores. [2] a highly customizable analytics application that identifies neighborhood/city-level market trends and buying opportunities.

Teamed up with Lendonate to build an ETL (Extract, Transform, Load) of Non-Profit Financials from the IRS for every U.S. NGO from 2007 to present. Custom querying interfaces and financial strength models bring the data to life for internal executives and clients.