- Use sckit - learn and TensorFlow for applications in finance, that include Support Vector Regression with new methods developed specifically for the program, Deep Reinforcement Learning that used a ensemble “stack” to increase investment returns, providing others with guidance on project to consider.
- Modeling simple to complex system in using various methods (e.g. Machine Learning, Deep Learning, Quantitative Analysis, Models developed from fundamental scientific approaches)
- Statistical Analysis of data used in model development and on predictions of models.
- Quantitative Analysis to develop models for trading.
- Trading Strategies.
- Stochastic Differential Equations.
- Volatility modeling for predictions and extended for trading.
- Risk analyst, mostly VAR and CVAR for complex portfolios - include Delta-Gamma.
- Options trading analysis for hedging portfolios.
- Portfolio modeling using Sharpe ratio and Omega ratio - which include risk factors.
- Factor modeling including Fama-French for portfolios.
- Interest Rate Models and Fixed income securities.
- Jobs involved with using Azure ML for machine learning problems - medical area and fraud detection.
- Programming Python, R and C++ for machine learning. My first job in Machine Learning was before the machine learning libraries existed so I programmer the Neural Network in C++ and the “front end” was in C#.
- PhD in Polymer Science & Engineering with Resume in Theoretical Polymer Physics - both computer simulations and mathematical modeling. Worked as an Assistant Professor in Theoretical Chemical Physics, expanding my ranging of fields covered.
- As with all projects in software field there has been work with SQL to get data that would be used the machine learning - alsoStored Procedures, Triggers, Views, Functions. The data normally needs to me modified (e.g. Normalization/Scaling and in many cases; dimension reduction and transformation of coordinate system (PCA, ICA, Nonlinear-Manifolds, Kernel-PCA, etc) before feeding the data into the model.
- Experience with Hadoop, pySpark (Big Data)
- Regular use of Visual Studio for C/C++, C#. SQL Server for database. NoSQL databases. Modeling using Excel and VBA. Applying high level mathematics to portfolio development and other models.
- Regular use on LINUX and programming in Python for Modeling and C++ to stay active with that language.
- Time-series modeling, AR, ARMA, ARMA, ARFIMA, GARCH Models, CARR Models. Support Vector Regression, Neural Networks, Decision Trees, Random Forest, Gradient Boost,etc.
- Bayesian statistics and models.
- Regression methods.
- Markov models.
- Monte Carlo methods including MCMC and simulations.
- Options modeling methods, analytic, numerical, and approximate.
- Optimization methods, linear, nonlinear. Analytic and numerical.
- Machine learning optimization and regularization methods.
- Portfolio optimization.
- Dealing with unbalanced data sets and incomplete data sets.
- More mentioned below (e.g. hyperparameter optimization, regularization, Gaussian Processes,etc. ).
- Monte Carlo Simulation Methods (e.g. Risk Analysis, MCMC, etc.), Fast Fourier Transforms, Stochastic Calculus, Numerical Optimization Methods such as, Adam, Gradient Descent methods, Nonlinear Conjugate Gradients, particle swarm, simulated annealing, etc.), hyperparameter optimization, Bias-Variance trade off, Statistical Modeling, Linear optimization, Mathematical Programming, Finite Difference Methods (with and without noise), and other numerical methods.
- Field theory, renormalization group, path integrals, Ito calculus, Wiener processes, Feynman-Kac, stochastic calculus, Fourier Series & Transforms, Laplace Transforms, Hilbert-Huang Transform, Wavelets, Filters, Partial Differential equations, Langevin equations, Special Functions, Functional Integration, Matrix Methods, calculus of variations, etc.
- Probability Analysis/Theory (including Markov processes, Bayesian probability). Copulas. Matrix Factorization. Statistical Analysis methods (e.g. PCA, ICA, CCA, LDA, SSA, M-SSA, Kernel-PCA etc.).
- Neural Networks (including Convolution Neural Networks, LSTM, RNN), Bayesian Neural Networks, Restricted Boltzmann Machines, Support Vector Machines & Regression, Gaussian Mixture Models, Ensemble Methods, Deep Reinforcement Learning and other Machine Learning methods (e.g. Decision Trees and Clustering--Supervised and Unsupervised, Random Forest, Gradient boosting, dealing with imbalanced data sets, e.g. SMOTE), NLP, Classifiers, SVM, SVR. Hidden Markov Models, Markov Switching Model. Loss Functions. Linear and Nonlinear Time Series Modeling (AR, ARMA, ARIMA, GARCH, CARR, etc.) - GMM, Yule-Walker & Box-Jenkins, Maximum Likelihood, EM, MCMC (Gibbs, Metropolis), Pattern recognition, regression methods (LS, WLS, GLS, NLS, PLS, SVR, Gaussian Process, Bayesian Regression). Factor models & multiple factor models.
- Q-Q plots, denoising methods (e.g. digital, wavelets, Fourier Transforms, Random Matrix Theory), Dickey-Fuller tests, Dixon's Q test and other Outlier Detection methods, statistics, scaling and normalization of data for use as inputs to models.
- Omega ratio method, Portfolio Optimization. Modern Portfolio Theory, Capital Asset Pricing Model, Arbitrage Pricing Theory, Black-Litterman Model, Shrinkage Theory, reduce risk (e.g. Global Minimum Variance) etc. Performance Measures (e.g. Sharpe Ratio), Random Matrix Theory. Trading Strategies. Risk analysis, CVAR and VAR (historical and Monte Carlo), Moody’s Risk Frontier. Econometrics Interest rate models and Fixed Income Securities. Volatility models. Options modeling ans solutions. Martingales. Ito calculus.
- C/C++ (including C++11 - g++), Python, C# .NET, SQL (SQL Server, MySQL, SQLite), Scikit-Learn, SciPy, Numpy, NTLK, TensorFlow, TensorForce, Azure Machine Learning, R, VBA, Julia, Sublime Text 3, Atom, gedit, Multithreaded programming (& Parallelization), Unix/Linux (Debian, Ubuntu & Cygwin), Hadoop/Spark/Hive (BigData, PySpark, Linux & Windows), Azure ML, AWS (to run python apps). ETL using SSIS, Data Warehouse, Data Mining. MVC, MVVM, MVP and other design patterns.
- OpenGL (and GLUT & GLUI) (OpenTK for C#), LINQ, Entity, WCF, SeviceStack, REST, Generics, WinForms.
- Mathematica, WireShark, Visual Studio, Excel, Tableau.
- CVS, TFS, GitHub.
- 20+ years of Scientific and Commercial Problem Solving; includes creating/solving Mathematical Models of complex physical systems and if necessary, Translating Models to Computational Solutions). C/C++11, Python, FORTRAN, C# .Net, Java. Experience with Network Programming (TCP/IP & UDP/IP) in C/C++, C# (Client/Server Applications).
Confidential, Miami, FL
Officer for Data Science and Quantitative Analysis
- I have been helping this group to obtain funding so they can go forward in their business - the help has been without pay because I believe in what they are working on.
- I have used Machine Learning and Quantitative Analytics to do some calculations that would aid their business objective documents for funding.
- I have also helped with reviewing documents, making suggestions for changes, design of their database and design of software user interfaces. The initial focus was to provide people, early in their careers to pay off their student loans with much more reasonable interest rate. The difficulty for these people is a lack of a financial history that could give them lower interest rate.
- The group ran into a problem recently with a number of potential presidential candidates and their plan for eliminating student loans, some of which involve more taxes. One of my early occupations was as an assistant professor and I could see problems that could be generated with that approach to eliminating student loans.
- Because of this, the group is now moving toward not just helping people get their student loans pair of but to also use their business model to allow for loans to people early in their careers to obtain other types of loans at reasonable rates.
- I 'm still helping them when I have time and hope they get their business moving. A bit more on their initial business objective with student loans can be found at www.READEnterprises.org .
Confidential, Greater Los Angeles Area, CA
Senior Quantitative Engineer
- Quantitative Analytics, Data Science/Machine learning, Risk Modeling and knowledge source for company with a focus on cryptocurrency trading.
- Position also involves Python, C++11 (some), SQL, and Big Data.
- Project I refer to as "using complex kernels in support kernel regression" which uses the properties of kernels to create the “complex kernels” (numerous of them). This was to allow for wider ranges of time for predictions using SVR. Also worked on back testing and denoising data also.
- Worked on the prediction methods for volatility of cryptocurrencies using a number of methods and models. I had an intern apply the “complex kernels SVR” to predict the 30-day volatility. Will be compared to CARR, CARRX, ACARR, GARCH, eGARCH and GJR-GARCH results. Also will be using Markov Switching for volatility to identify regimes in volatility. SVR with complex Kernels will be used to look at historical volatility.
- Completed the first phase of Deep Reinforcement Learning for portfolio management --models completed that involve CNN, RNN, LSTM, & Dense Networks where Tensorflow used.
- Completed implied volatility predictions (and predicting the volatility surface) using the SVR and complex Kernels.
- Identify the tops/bubbles in cryptocurrency markets and CRIX index.
- Worked with others to help them with statistical methods (e.g. pair Trading using Cointegration & also Copula approach) for Cryptocurrencies Trading.
Confidential, Boston, MA
Data Scientist & Quantitative Analyst
- Stock selection based on machine learning methods.
- C++ for Support Vector Regression completed.
- One using an Ensemble Method (dimension reduction followed by neural network for return predictions).
- Second involves testing my own kernel function for trading.
- These done in python.
- Used by the company for consulting on issues ranging from portfolio optimization to issues with other developer’s machine learning models and supplying ideas for models to pursue which I do not have time to work on.
- Part of review group on a person's dynamic mode decomposition (DMD) modeling for day trading.
Confidential, Cincinnati, OH
Senior Data Scientist
- Initial project, apply Machine learning to fraud detection and traditional methods for fraud detection, using Hadoop/Spark/Hive (BigData), Python, Scikit-Learn, MLlib, C++11/14.
- Applied Machine learning to fraud detection using Hadoop/Spark/Hive (BigData), Python, Scikit-Learn, MLlib, C++.
- Initial project focused on anomaly detection (which includes fraud) in sales data. The first four programs in python used variations of an unsupervised methods that I created - not at allowed to give details currently. The model will be modified in the future to make is semi-supervised.
- Other methods used focused on unsupervised learning (e.g. SOM and K-Means) which is good for finding current anomaly methods and identifying new anomaly methods while supervised methods are good for identifying current anomaly methods.
- Support vector machines and Random Forest will also be involved with the work. The idea was to use multiple methods on identifying anomalies with these multiple methods analogous to fighting a war with layers of troops. An additional method, variation of Benford’s Law that also uses machine learning was investigate.
- I suggested that the company move to using Flink in place of Spark since it allows for true real-time streaming (online vs batch) of data which would benefit the company in a number of areas. Created a Tweepy application (using the NTLK library) that would be used to find a company's coupons posted on twitter.
Confidential, Aventura, FL
- This start-up company is based on the research that discovered a way to measure the temperature of the brain (different from measured body temperature) using a simple sensor oriented in two locations on a person's head. Measurements can be performed in various states (e.g. sleep, exercise, etc.) as a function of time.
- The temperature vs time may be linked to health conditions of the brain & body -- Alzheimer's disease, pregnancy, just to name a few.
- I worked with the data collected to understand the consequence of different temperature profiles & also predictions at later times. Methods employed are 1) "cleaning and smoothing" data 2) creating machine learning (ML) models for Pattern Recognition and Time Series predictions. ML approaches include using Azure Machine Learning, R, and C++11 programming.
- C++ is introduced into the Azure ML via the RCPP package (R) to provide fast ML & also models currently not available in Azure ML can be used by via C++ -- starting with version 0.10.3 of RCPP C++11 can be used. Models to be investigated for Pattern Recognition & Time Series include Neural Networks, Bayesian Neural Networks, Support Vector Machines (and Regression), Random Forests, Correlation methods etc.
- Frequency analysis, PCA, & ICA analysis of data. There will always be the search for the "best" ML to identify specific temperature patterns for the various health conditions and create ML to predict possible future conditions. Developed an unsupervised method (SVM.
- Neural Networks, k-means were also used for comparison) to identify patterns in brain temperature, time-series - could be applied to other types of time-series. Completed an Azure Machine Learning program for Pattern Recognition, built the Web Service to allow access to it, and gave the "client application group" the code to access the web service.
Confidential, Dania, FL
Sr. Software Developer
- C# .Net, SQL Server (queries, database updates, stored procedures), ASP.Net (WebForms and MVC), JQuery Widgets, Entity Framework (simple queries and database insertions along with LINQ to Entity queries) developing and maintaining web based applications that are used “in house” and by clients (referred to as “Agents” within the company).
- I have been working with Marmalade C++ 11 development environment, which allows the development of cross platform applications (and games) on mobile devices.
Confidential, Boca Raton, FL
Sr. Software Developer
- Software development for Medical Supplies & Pharmaceuticals company C/C++, C# .Net, SQL Server, Python, RavenDB (NoSQL), InfluxDB (NoSQL), SeviceStack (Web Services using REST and extracting data from JSON objects), Ext JS (from Sencha for UI), HTML5, GitHub (Bug Tracking, Application Repository).
- The work starting with indexes (clustered and non-cluster using C++) for quick searches, and development in C/C++ implementing the Levenshtein Automaton algorithm.
- The Levenshtein Automata algorithm was used for fast fuzzy searches - it runs faster than other algorithms tested and code will run on both Windows and Linux operating systems. In addition, I also developed c code for AGREP (which could also be used for “fuzzy searches”). This will be used as a comparison to the Levenshtein Automata.
Confidential, Fort Lauderdale, FL
Sr. Software Engineer
- C# .Net WinForms (all other application mentioned heard, unless noted where WinForms apps) application to extracted data from a DB using ADO.NET (there were stored procedures involved). Note: Started with plans on using Entity framework; however, DB was not created properly so use of Entity (or LINQ) was not possible. Statistical analysis of data in application; then the saving some data requested by a senior chemist in and Excel worksheet (programmatically). Note: One result of application was discovery of a manufacturing issue which involved the diabetes meter test strips.
- Two SSIS projects that took data from excel files - data was from test on people around the country to provide a wide range of data for analysis. Modified the data and added it to the DB. Prior to working the Monte Carlo project, the data needed for the project was added to the database.
- Project that create a histogram (Hematocrit vs Glucose) and determined a valid the probability distribution from data in. Note: Test to see determine “best” probability distribution employed “qq plots.” Note: Displays of data in application used OpenGL (via OpenTK library)
- From the probability distribution, I was working on developing a Monte Carlo simulation application that would be used for Risk Analysis. Database design associated with project listed above
Confidential, Deerfield Beach, FL
Sr. Software Developer
- Development and maintain ASP.NET/C# Web applications (Administrative Apps, Websites, and others including controls/ASCX) for a SMS texting system which was used to send product messages to customers at regular intervals and messages to support product payments.
- Development of Service applications and DLLS in C# .NET to support the messaging system -- Manger of dates for sending messages, Use of Binds (SMPP) to send and receive messages, Listen and response to customers sending messages (MO).
- Creation of WCF Web Service for billing support with Verizon customers. On the Verizon project, the company was not sure if the Web Service was going to be SOAP based or a REST application so I developed concurrently for SOAP and REST since there were time limitations.
- Queries and stored procedures in SQL Server to support the messaging system and provide data to business department Data support and development within the messaging system involved LINQ, Enterprise Data Access, or ADO.NET
- Creation of a design document to replace the existing mobile messaging system and other documents to provide procedures and description of the existing system. This was for a partner company that Confidential had contracts with.
- WinForms applications to support product.
- Developed the Architecture for a new text messaging application.
- Python was used for scripting.
Confidential, Miami, FL
Senior Software Engineer
- Development of PC applications and embedded software for the installation, configuration, operation/ control and updating specialized audio/video systems. These applications utilize C# .Net, Borland C++, and Portable Embedded GUI (PEG and PEG+) development environments. Assisted with trouble shooting and feature addition on InstallShield projects. Developed Android (Java) applications.
- Firmware Update application for product systems (socket comm.), IR Capture Station (serial comm.) used to record, modify, store IR remote signals for use with company systems), Imaging Apps (GDI+), Automated Testing for Installs and Firmware Updates, Serial Comm. Diagnostics for iRemoteTS wireless touchpad.
- GUI Application to control system audio over Wi-Fi (socket communications).
- Created touchpad applications for company Contact, iRemoteTS (wireless ZigBee), TS-Pro.
- Auto IP configuration and Restore on PC (DLL), Icon resizing and color palette optimization (DLL), Registry Sweeper (clean-up) Utility, ICS Wireless Topology Tool to query the system components and display topology with network settings. COM applications for cross platform work.
Confidential, Boca Raton, FL
Senior Software Engineer
- Developed medical device software utilizing C# .Net, SQL Server and ADO.NET.
- This involved multi-threaded programming, inter-process signaling, GDI+, GUI design, testing and debugging.
- I also contributed to the creation of software requirements and specifications.