Apptio, an IBM Company

Staff AI/ML Software Engineer

Phenom

In my most recent job role, at Phenom, I led cross functional teams to build and optimize entity matching machine learning pipelines to match candidate profiles with 3rd party profiles for profile enhancement.  I architected and led creation of entity matching application's underlying infrastructure (with NLP).  I modeled maximum matching percentage to identify the filtering step was the major pain point of the application preventing 30% of profiles from entering the identification pipeline.
Key Accomplishments:
- Optimized existing ML model for entity matching by ~20% by employing machine learning tools, such as random forest, auto-encoder, and Siamese neural networks
- Improved data labeling pipeline to allow for thousands of labeled comparison enabling consistent model monitoring.
- Led team on including machine learning into fit score (profile - job comparison) resulting in a ~5% performance increase.
- Created metrics for measuring data quality.
- Consulted with AI teams across the company to improve data science and ML techniques and applications.

Principal Machine Learning Engineer

TIBCO Software Inc.

While I was at TIBCO I served as product lead of ML projects within the integration division of the TIBCO. I enabled model deployment into event flow applications by integrating TensorFlow into Project Flogo (flogo.io) a platform for building event driven tools that when coupled with my design and implementation of the CatalystML Specification to ensure data transformations from development to deployment in allowed no-code deployment of apis. I built ML applications and models at production, proof-of-concept, and demo levels.  I also built AWS CloudFormation and Lambda infrastructure for production ML products.
Key Accomplishments:
 - Enabled model deployment into event flow applications by adding Machine Learning and AI to open-source golang framework Project Flogo (flogo.io).
-  Developed CatalystML specification to describe data transformations to allow easy deployment of machine learning pipelines to production.
 - Created lambda based continuous training application for matching columns between databases (i.e. Salesforce to Marketo).
 - Used Cloud Formation to productionalize the data pipeline and lambda structure of the above entity matching product.

Machine Learning Engineer | Architect

Narvar

Ensured smooth workflow by preparing designs for new data science and machine learning centric products. Created and installed improved supply chain prediction system, while erecting training set of billions of data points distributed across multiple backend systems (SQL and NoSQL dbs) and third-party APIs. Established engine for organizing millions of consumer comments to attain in-depth knowledge of e-commerce to consumer supply chain pain points. Deployed NLP through CNNs in Tensorflow with GloVe vectors for comment classification.
Key Accomplishments:
- Created data products that removed major pain points of the core functionality of consumers post-purchase online experience for our clients.
- Utilized comment classification in Tensorflow for developing an engine for categorizing millions of user comments in order to comprehend pain points in e-commerce to consumer supply chain.
- Created engine for classifying millions of consumer comments to understand e-commerce to consumer supply chain pain points.
- Established training set of billions of data points spread over numerous backend systems (SQL and NoSQL dbs) and third-party APIs; resulted in increasing supply chain prediction system.

Data Scientist

Slice Technologies, Inc

Created a nonlinear global optimization algorithm written in Python from scratch. Led and collaborated with contractors on construction and co-building of AWS infrastructure through utilization of EC2, docker containerization, auto-scaling, and ECS. Acted as an integral member of the diversity working group.
- Responsible for statistical data accuracy of point-of-sale data for all clients driving millions in revenue.
- Created data products by engaging with other fellow members.
- Boosted morale by training a new team, while maintaining team output
after every other team member left the group.

Insight Data Science

While a fellow at Insight, I developed Wiki-tics and political dashboard, using Wikipedia page views. I actively gathered and analyzed 2-years of Wikipedia page views in Python. Moreover, I employed anomaly detection to identify trending topics. I detected political topics using Java, while extending Wiki-tics.
Key Accomplishments:
 - Developed data product from conception to hosting EC2 web page for outlier detection through utilization of machine learning tools.
 - Created data products by engaging with other fellow members.
 - Enhanced data science and machine learning skill set for industry employment.

Fellow | Data Science

Independant Researcher

- Developed machine learning (Random Forest, ADA) python code to predict temperature in Pittsburgh.
 - Investigated “Predict survival on the Titanic“, “Bag of Words Meets Bag of Popcorn”, “Poker Rule Induction” challenges at Kaggle.com
 - Collaborated with theoretical chemistry group at University of Groningen studying organic solar cells.
 - Continued development of materials prediction of materials with numerical optimization methods.

Binghamton University

- Predicted existence of new materials using numeric optimizations (genetic algorithms and conjugate gradient) in C and noSQL databases.
 - Cluster administrator (deployed cluster and setup HPC pipeline) for 192 core machine.
 - Co-Developed MAISE (Module for Ab Initio Structure Evolution), http://maise-guide.org, a novel materials prediction software.  
 - Investigated metal-borides and transition metal oxides with ab initio and evolutionary search techniques.

Research Associate

École nationale supérieure de chimie de Paris

- Investigated corrosion of stainless steels in nuclear reactors by constructing computational experiments.  
 - Modeled corrosion dynamics by developing kinetic Monte Carlo programs in C.

Post-doctoral researcher

LETI

- Designed computational experiments to drive development of faster transistors.
 - Created a pipeline for large numbers of calculations with Python, Fortran, and Bash scripts.

Post-Doc

I build machine learning models and pipelines to ensure product and client success (internal or external).

Contact me if you are interested in someone with an interdisciplinary background (PhD 2009) or if you just want to chat about Tech, Machine Learning, AI, Life, the Universe, and everything.

My interdisciplinary background plus my 8+ years in industry means I an extremely suited to solve problems by building and leading teams to build effective products and features for customers while eliminating pain points for internal data and model pipelines decreasing costs and increasing efficiency.

I specialize in:
- Taking large ambiguous problems and producing and deploying solutions
- Data Science, Machine Learning, and Software Engineering
- NLP, CV, Entity Matching, Outlier Detection, etc.
- Data cleaning, processing, and transforming for analyses and machine learning that exceed expectations

Taplio

Abram Van Der Geest, Ph.D.'s Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Abram Van Der Geest, Ph.D.'s Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Abram Van Der Geest, Ph.D.'s Linkedin Analytics