Provided scientific, technical, and organizational leadership on cross-functional customer delivery teams.-- Hotel multi-horizon demand forecasting --Led a team of 2 data scientists, 1 MLE, 1 security architect to build an MVP for 4 demand forecasting tasks that would support the customer’s in-house revenue management system for their properties in North America.

Engagement Lead on 5 of 6 customer-facing projects and with Novartis (Science IC), Nationwide, BP p.l.c., Bridgestone, and Best Western. Use cases spanned NLP, forecasting, signal processing, computer vision, and price optimization.-- Coefficient of Friction Prediction --Led a team of 2 data scientists to build a POC for real-time estimation of the coefficient of a friction from a pair of 20kHz accelerometers fitted to the inner wall of a tire.-- Fuel Demand Forecasting and Price Optimization --Led a team of 5 data scientists to build an MVP of fuel site/product level demand forecasting and pricing optimization for the customer’s fuel sites across North America.-- Product Competitive Intelligence -- Led a team of 1 data scientist and 1 MLE to build an MVP of a solution to predict when, by how much, and in which direction the customer’s competitors would adjust the interest rate (cap rate) of their fixed indexed annuity products over the next year.-- Adverse Event Detection -- IC science contributor on an engagement to automate adverse event detection at scale in a multilingual, low resource setting for the purpose of mitigating regulatory risk.

Responsibilities include deep learning, evaluation of alternative data sources, and development of new products and enhancements to the current product offering- Design and implement deep learning pipelines. 
Create machine learning pipelines to scale current processes and coverage (Scala/Spark, Python/TensorFlow)- Developed a scalable nearest neighbor solution for identifying merchants in billions of transaction descriptions using a metric space data structure. (Scala, Spark)- Mathematical and statistical modeling. 
Documented current estimation methodology, derived sensitivity analysis, and provided mathematical intuition for the impact of attrition and other biases. (Pen, Paper, Latex)- Spearhead development of new alpha-generating and macro-nowcasting services
. Evaluate large alternative-data sources to reveal strengths/weakness and identify the best strategic partnerships. - Built the R&D EMR evaluation stack. 
Built the R&D team’s EMR stack including boot actions and other configuration scripts that allow myself and other data scientists to carry out full evaluation pipelines interactively in notebooks on EMR. (Spark/Scala, Zeppelin, EMR, S3)

Responsibilities include developing machine learning infrastructure and algorithms for forecasting macroeconomic conditions from high dimensional datasets composed of millions of real-time and historical financial time series.- Designed and developed novel persistent time series data structure. Enables the storage of 25TB of time series vintages and retrieval at sub-millisecond speeds, replacing 1980s data infrastructure based on FAME time series database. (Java, JNA, C, PostgreSQL)- Designed and Developed REST API for managing forecasting models. The API negotiates batch submissions to a 50 node SLURM HPC cluster, making it possible to build and benchmark models from a webpage and trace errors back to the exact system, job, and model configuration that launched each model. (Python, SLURM HPC, R, Matlab)

Responsibilities include supporting the software and data infrastructure needs of Board policy and academic research in the area of systemic risk, complexity, and liquidity. - Published in Top 10 Finance journal using natural language processing and causal inference. 
Research included surveying NLP literature to develop a measure of sentiment expressing uncertainty. (R, Python)- Designed and developed entity resolution algorithm for scoring input to a knowledge graph. 
Designed with scalability and generality in mind, this package is also used by Research Assistants to link companies across datasets without common identifiers—to task that used to be custom-coded for each project. (Python)- Independently developed a network analysis approach to systemic risk surveillance. 
Originally published on federalreserve.gov, this methodology is now regularly featured in the confidential Quantitative Surveillance report presented to executives as the Federal Reserve. (Python, R, Gephi)- Designed and led multiple courses in R, Python, Hadoop, and Spark.
Includes lecturing for a graduate-level econometrics course taught in R and hosted at the Board for Howard University students. (R, Python, HDFS, Hive, Pig, Spark)- Developed Hive connector for Stata.
Enables PhD Economists with little technical training to take advantage of big data technology. The only plugin that has been developed to allow Stata to access Hadoop. (Java, Hive, Stata)- Developed an internal knowledge base to retain institutional memory. 
The site serves as a hub for fixes to problems encountered around the Federal Reserve. (PHP, MySQL, D3, Kerberos)

Responsibilities included supporting academic research and training junior Research Assistants.- Project and data management for policy and academic research. 
Assistance led to 5 publications in Top 10 Economics journals.- Prepared briefings for executive stakeholders, Governors Stein and Powell. 
Ensured accuracy of data analysis and clarity of visual presentation. (R, SAS, LaTeX, Word)

- Advised a class of 140+ 300-level, undergraduate, Game Theory students- Led weekly homework and exam help sessions- Verified game theoretical assignment and exam solution sets- Graded and corrected student homework and exams

Responsibilities included project management, design, and development of 40+ web sites.