Projects
- Python
- Tensorflow
- Keras
- OpenCV
Convolutional Neural Network
Developed a Convolutional Neural Network using the Residual Network (ResNet) architecture that detected if an MRI
brain scan had a tumor or not with an accuracy of 97 % and a recall of 94.4%. Radiologists may minimize the risk of missing brain tumors (FALSE NEGATIVES) with the aid of this model as a second pair of eyes.
Dataset: HERE
Dataset: HERE
- Tensorflow
- Python
Wide and Deep Network
Developed a Wide and Deep Neural Network (WDN) to predict if a song was going to be a hit or a flop using a Spotify Music data set. These insights are of interest to musicians who might need prior knowledge in order to properly allocate resources for marketing their songs. The final model had a precision of 75%.
Dataset: HERE
Dataset: HERE
Galaga |Reinforcement Learning
Developed the game of Galaga and an AI agent to play against humans.
Developed in a team of 5 using the Sequential or Waterfall Software Development Process and Model View Controller (MVC) for the architectural pattern, (Model = backend or Game Engine, View = screen, Controller = player or AI agent)
Developed in a team of 5 using the Sequential or Waterfall Software Development Process and Model View Controller (MVC) for the architectural pattern, (Model = backend or Game Engine, View = screen, Controller = player or AI agent)
C++
Covid 19 Search Engine
Search engine built in C++ to help biologists, virologists and other scientists to efficiently find already-published research articles about COVID19 using a dataset consisting of thousands of scientific scholarly publications from Allen Institute for AI in partnership with Chan Zuckerberg Initiative, Microsoft Research and several other organizations. More detail:
HERE
- Java
- HTML
- CSS
Pick Me Up
This is an automated delivery system for in-store pick up that we built using Java and a Postmates API. Imagine when you buy commodities at Walmart and you can't collect them due to traffic, time constraints or other commitments. PickMeUp finds someone online who can deliver it at your doorstep.
- Java
- ReactJs
Stock Ticker
STOCK DATA VISUALIZATION App completed in a day long hack-a-thon style at Goldman Sachs Engineering Essentials Summit in New York City by a team of 5. It comprised of a Java Back-end and a React Front-end. At the end we presented to the senior engineering / Technology leadership of the firm.
GS's Intellectual property laws prohibit code sharing
C++
Search Algorithms
Implemented Dijkstra, A*, DFS & BFS search algorithms to find a path between 2 given nodes & return a path from source to destination node with total cost incurred. The search is perfomed over both an adjacency list and adjacency matrix forms of loaded graphs. Strategy Design Pattern was employed that allows the selection of an algorithm at runtime.
C++
Dynamic Programming
Applied Dynamic Programming & Naive Brute-Force techniques to solve the Travelling Salesman's Problem (TSP) by finding a Hamiltonian circuit for a given list of nodes and positions on a graph. Shortest path is returned. (OPTIMAL). Factory Pattern was employed to allow the creation of objects without knowledge of the class of objects that would be created.
C++
Heuristic & Metaheuristic Algorithms
Applied Genetic Algorithms & Simulated Annealing techniques to solve the Travelling Salesman problems by finding a Hamilitonian circuit for a given list of nodes and postions on a graph.Algorithm returns the shortest (OPTIMAL). Factory Strategy pattern was employed to allow the creation of objects without knowledge of the class of objects that would be created.
Contact