FPL Optimization Modeling
I have implemented, improved, and shared an optimization model for Fantasy Premier League (FPL). I regularly publish the building steps in a tutorial series on YouTube, and show how to apply analytics to fantasy sports. I have covered topics like mixed-integer linear optimization, row generation, multi-objective optimization, stochastic optimization, simulation, and sensitivity analysis.
◹ Always Cheating podcast interview
I have built the first and only open-source website focused on Fantasy Premier League (FPL). The website provides a variety of analytical tools, including descriptive, diagnostic, predictive, and prescriptive analytics. This one-of-a-kind website promotes using optimization and analytics in everyday problems and draws ~200 daily users.
Expected Goals (xG) Modeling for Rocket League
Optimization for Expected Threat
Expected Threat (xT) model measures the impact of ball movements in soccer, such as shot, pass, and dribble. I have proposed two optimization models to overcome symmetricity and consistency issues of the xT model. The model can be easily applied as a wrapper after summarizing ball movements, and provides a better accuracy.
Venue Optimization (SAS Institute)
I have designed and produced a Venue Optimization tool for venue managers to safely admit fans into seated venues under COVID-19 restrictions. The optimization model I developed uses state-of-the-art optimization solvers and modeling tricks involving decomposition and maximal clique generation. The tool has been covered in more than 100+ local news all around the world, covered by WRAL in a video interview, a team member is interviewed by Forbes, and the work has been mentioned in high profile journals. The model is also used by NFL for the Super Bowl in 2021.
Seating Optimization Game (SAS Institute)
sasoptpy: Python modeling interface (SAS Institute)
I have developed and maintained the Python interface for SAS Optimization solvers, sasoptpy. This Python package is capable of taking optimization models in native Python format and can be used to describe both linear and nonlinear optimization problems, as well as support concrete and abstract modeling. It fully integrates with SAS Optimization API and can be used to solve optimization problems in cloud with familiar syntax for developers.
Traveling Euro Fan Problem
Due to the unique format of Euro 2020, we have used the traveling salesman problem within Euro setting to find a way to visit all 12 stadiums while catching an international game in each. We have used several modes of transportation data and benefitted from the parallelization to quickly generate optimal paths. The source code is shared on GitHub
Traveling Baseball Fan Problem (TBFP)
I have implemented the TBFP in Python and have shown the optimal solution of visiting all Major League Baseball stadiums was possible in 24 days and 3 hours in 2018. I have shared the model and the implementation details on GitHub
Hyperparameter Optimization for AutoML (SAS Institute)
I worked with AutoML group at SAS and implemented the Bayesian Optimization tool for Hyperparameter Optimization toolkit within AutoML. The implementation can help Machine Learning users to figure out parameters to maximize the training score on their test sets by suggesting ideal parameters in an iterative search method. 'Bayes' is available under AutoML as an optimization method.
Operations Research Specialist
I work under Analytics Applications group within Advaned Analytics division at SAS Institute R&D. My responsibilities include developing Python interfaces to SAS Optimization solvers, working with customers to transform their business problems into mathematical models, and providing support to Center of Excellence Optimization group for their consulting engagements.
I have worked as a teaching and research assistant during my PhD at Lehigh Industrial and Systems Engineering department, for both Engineering and Management schools.
PhD in Industrial Engineering
Dissertation: Exploiting Structures in Mixed-Integer Second-Order Cone Optimization Problems for Branch-and-Conic-Cut Algorithms
MS in Industrial Engineering
Thesis: Final Phase Inventory Management of Spare Parts under Nonhomogeneous Poisson Demand Rate
BS in Industrial Engineering
Optimization Modeling with Python and SAS Viya
SAS Global Forum (2018)
The first heuristic specifically for mixed-integer second-order cone optimization
Optimization Online E-Print ID (2018): 01-6428
Warm-start of interior point methods for second order cone optimization via rounding over optimal Jordan frames
Optimization Online E-Print ID (2017): 05-5998
Effects of disjunctive conic cuts within a branch and conic cut algorithm to solve asset allocation problems
Lehigh ISE Technical Report 18T–002 (2018)