Andrew Fast

Andrew Fast

Data Scientist

About Me

I am a creator, communicator, and practitioner of advanced algorithms for understanding complex data sets. I have worked as a researcher developing new algorithms for data science, a software engineer deploying robust systems, a data scientist applying advanced algorithms to solve a specific business problem, and a team leader and manager enabling the delivery of analytics solutions at high-level within business constraints.

I am currently the Chief Data Scientist at CounterFlow AI, a network security company focused on network forensics and threat hunting. Previously, I held the position of Chief Scientist at Elder Research, Inc.. I earned my MS and Ph.D. degrees in Computer Science from the University of Massachusetts Amherst, specializing in algorithms for causal data mining, and for analyzing complex relational data such as social networks. With Dr. John Elder and four others, I co-authored a book titled Practical Text Mining that was published by Elsevier in 2012 and won the PROSE Award for top book in the field of Computing and Information Sciences for that year.

Areas of Expertise

Data Science

Telling the truth with data

Data Science

Data science is all about extracting meaningful knowledge from data and effectively communicating that knowledge to others. First and foremost, meaningful knowledge must not arise from spurious correlation or be displayed in misleading ways. Second, it must be repeatable and arise from a consistent process. In pursuit of these goals, two major themes of my work are understanding how learning algorithms perform in different circumstances and extending traditional data mining approaches to the areas of text, complex networks and causality.

I have worked in wide variety of industries and application areas including consumer product goods, healthcare, non-profits, marketing and insurance. I have helped hundreds of organizations ranging from large multi-national corporations to single person start-ups to build consistent, performant data science systems.

Network Graph Image

Graph Analytics

Though most data science approaches assume independent instances, data rarely are truly independent. Graph Analytics allows you to easily reason about and take advantage of dependencies between data instances. I have experienced the benefits of graph analytics in applications ranging from fighting fraud to understanding coaching success in the NFL.

Text Analytics

Text Analytics

Having engineered systems for processing text including NLP engines, text classifiers, dynamic search engines and other tools, I've found the key to having success with text analytics is using a combination of both linguistic and statistical reasoning. In addition to building text analytics software, I've authored a text book on text mining and routinely teach a day-long seminars on text analytics concepts.

Causation Graphic

Causation and Experimental Design

Moving from correlation to causation is one of the current frontiers of machine learning research. I've contributed to multiple algorithms to identify causal knowledge from observational knowledge alone including my thesis work on improving the accuracy of structure learning algorithms for Bayesian networks and work with automated detection of quasi-experimental design.

Data Science Strategy

Data Science Strategy

I have had the privilege of working with hundreds of organizations to implement data science, analytics and/or machine learning techniques in their business environment. These experiences have shaped my understanding of successful deployment of data science models in a variety of settings. Specific experience includes starting an Analytics Center of Excellence, hiring data scientists, acquiring analytics software, and assessing analytics maturity for organizations ranging in size from start-ups to large multi-national corporations.

Other Interests

NFL Coaching Trees

NFL Coaching Trees provide a unique opportunity to quantitatively explore how influence propagates among individuals. The best coaches produce many assistants who go on to be head coaches, but these assistants are not always successful. I hope to understand why. See more at:

nflcoachingtrees.com

Data Science and Functional Medicine

After wrestling with chronic illness in my family for the last several years, I am learning that many unexplained medical conditions such as chronic fatigue or food sensitivities can arise from the interaction of multiple factors including nutritional deficiencies, genetic predispositions, and environmental toxins. This is an ideal situation to use machine learning techniques to help identify correlations between weak signals. I'm interested in collaborating with others in this area.

Contact me at a...t@gmail.com

Swarm and Emergent Behaviors

The behavior of complex systems and how seemingly complex behavior emerges from very simple systems intrigues me. I wrote a thesis prospectus for the AAAI Doctoral Consortium describing how graph analytics and machine learning could be applied to better understand the emergence of complex adaptive systems.

Learning Models of Macrobehaviors in Complex Adaptive Systems