Data Scientist2019 - PRESENTTactical Foresight Consulting, LLC
- Used Python and R for data collection and statistical modeling, leveraging unsupervised models when labeled data was scarce.
- Determined and designed technological capabilities, showcasing proof-of-concept (POC) of said capabilities to the client.
- Created D3.js and Tableau visualizations for clients which reported needs.
- Built a program to parse court documents to count reference to legislative statues and detect novel combinations of laws.
- Used Bayesian Networks to visualize the influencers of a ballot measure pass rate.
- Used NLP to create a graph of activities from scraped data from news articles.
Data Scientist (Consultant)2018 - 2018MatchPoint
Technologies: Python, SQL, Regex
- Suggested, created, and tested a framework of unsupervised methods to detect suggested suppliers.
- Presented results in a clear manner and developed flowcharts of how the system works.
- Used natural language processing dependency trees to create categorizes as a training set.
- Extracted useful search features from the text, created classifications for matching and search problems, and worked on experiments which resulted in successful unsupervised matching algorithm with approximately 96% accuracy.
- Developed metaheuristics for creating and sourcing training datasets.
Data Scientist2017 - 2018Systematrix Solutions
- Used Spark MLlib via PySpark for outlier detection on GraphX RDDs.
- Presented and coded new algorithms for graph analytics using GraphX and Scala.
- Used PySpark for fraud analytics on banking records via RDD transformations, filters, and joins.
- Created, modified, and benchmarked machine-learning algorithms for statistical inference on network properties and money laundering prediction in a Docker container.
- Routinely provided qualitative insights into upcoming roadblocks to meeting projects and customers needs before it was a noticeable problem.
- Took the initiative to develop and present data privacy policies, standards, processes, and local and international legal requirements.
- Translated the fraud investigators' goals to extract essential subgraphs via graph-properties filters and transversals that delivered explicitly fraudulent connections in addition to causing a reduction processing time for analytics.
- Prescribed a strategic approach to handle changing algorithmic regulations, burst-out-fraud, and take-over-fraud.
Operational Intelligence Analyst2015 - 2017Stanford University
Technologies: Python, R, Neo4j, SQL, MongoDB, Tableau
- Used mathematical techniques and fitted statistical models to analyze data related to business problems and visualized the results in Tableau dashboards and Neo4j.
- Visualized and Identified contextual data that was needed, patterns, summary statistics and trends using (but not limited to): graph analytics, non-parametric ensemble models, Bayesian inference, and natural language processing (NLP).
- Adjusted the code for multicore parallel processing on computer clusters and used MapReduce functions to aggregate data for customer profile to supplement Neo4j database.
- Used Cypher (Neo4j QL) to add features such as fund amount to graph database of transactions.
- Automated a system to categorize any text using an unsupervised model that eliminated the need for manually finding cluster centers or reducing the time to find density parameters.
- Leveraged glove vectors (or Word2Vec) to classify an activity's risk which was extracted from text using NLP and then modeled their impact as a network/graph.
- Constructed statistical frameworks and code by utilizing new machine learning programs; I then presented them at conferences and expos.
- Met with clients and listened to their needs in order to design solutions to those needs.
- Visualized the above-mentioned data in a Tableau dashboard.
- Collaborated on multiple high-priority projects and made key contributions to the team’s long term strategy meetings.
- Solved problems with a user-friendly explanation of the methodology and with minimal oversight.