Schlagwörter
Skills
- machine learning
- time-series forecasting
- quantitative finance
- optimization
- cloud computing
- using data to answer complex research questions
- coordinating freelancer teams in an international environment
- linear models: OLS, WLS, GLS, Lasso and Ridge Regression
- nonlinear models: Random Forest (classification/regression), Gradient Boosted Models, Gaussian Process Regression, Feed-Forward Neural Networks, Kalman Filters
- R and Python: daily experience since 2011 and 2016, respectively
- SQL: exposure to Microsoft SQL Server, SQLite
- JavaScript: created web dashboards via jQuery and DataTables
- C/C++: used for Arduino projects; sometimes also via Rcpp
- LaTeX: employed for research reports and documentation (mostly via knitr)
- R: h2o, (d)plyr, data.table, ggplot2, zoo, lubridate, Rcpp, knitr, roxygen2, testthat
- Python: h2o, pandas, numpy, scikit-learn, keras, sqlite3, flask, boto3, requests, unittest
- AWS (EC2, S3)
- Git, SVN
- Docker
Projekthistorie
- Created machine learning models for energy load forecasting (data exploration, model prototypes, hyper parameter optimization, documentation)
- Software maintenance for past projects
Selected projects:
- Created cloud-based hyper parameter optimization infrastructure supporting Bayesian, random and grid search in Python
-
Developed transaction cost model for global futures markets
-
Worked on time-series cross validation methods for automated hyper parameter optimization
Selected projects:
- Developed Random Forest variable selection method for low signal-to-noise ratio data
-
Implemented different weighting schemes for Random Forest
-
Created prototype models (gradient boosted trees and neural networks) for time-series forecasting
Selected projects:
- Follow-up project on cloud migration: developed an API interfacing the Amazon cloud (AWS EC2); the API handles the client’s custom requirements regarding logging, cluster management and per-user cost attribution
-
Developed R and Python backends for machine learning model training
-
Implemented hyper-parameter optimization infrastructure in R
-
Initiated, led and worked on a project for time-series prediction with Random Forests; the solution is used by the client (a trading firm) for their main trading strategy
-
Applied statistical methods to detect structural breaks in time-series data
Selected projects:
- Developed Random Forest and Neural Network models that generate return forecasts for various financial markets
-
Research on new predictors for financial markets
-
Proof-of-concept project to move machine learning infrastructure into the AWS cloud
Selected projects:
- Research on fundamental-data-based predictors for financial markets
-
Developed allocation algorithm based on dynamic programming
-
Developed cost model for ETFs