Experimenting with the QAOA on the TSP problem

Projects

Throughout this project I embarked on an exciting journey into quantum computing, and delved deep into solving the Travelling Salesman Problem (TSP). Blending solid research with creative thinking, I developed methods for solving the TSP using the Quantum Approximate Optimization Algorithm (QAOA).

To do this I transformed the TSP into the Quadratic Unconstrained Binary Optimization (QUBO) model using a few different methods. Then, I tested these models on quantum simulators to see if they would actually work. The highlight was running a real TSP instance on one of IBM's Quantum Computers, proving the potential of quantum computing in solving complex problems.

QUBO Formulation of the Cardinality Constrained Portfolio Optimisation Problem

A large number of optimization problems can be formulated through the Quadratic Unconstrained Binary Optimization (QUBO) model. This approach enables a singular heuristic solver to tackle a multitude of problems, allowing companies to concentrate their investments on the development of a single unified solution. Notable examples of companies using this model include Fujitsu, with their Digital Annealer, and D-Wave, working on their Quantum Annealer. The QUBO model also closely resembles the Ising Spin model which allows it to extend its versatility to solving problems using quantum systems.

This project honed in on the development of an algorithm dedicated to transforming the Cardinality Constrained Portfolio Optimization (CCPO) problem into a QUBO. Subsequently, a series of test CCPO problems were solved. The CCPO problem revolves around identifying the optimal portfolio from a pool of X assets, where the portfolio's asset count must be less than X. The project emerged as highly successful, demonstrating that solving the resulting QUBOs unveils the efficient frontier of portfolios.

Further developement of the project, enabled the QUBOs produced to be solved using Fujitsu's Multi-Objective Digital Annealer.

Econ Archive

Econ Plaza is a web-based social media platform curated by economists for economists. This project involved adding additional functionality to Econ Plaza through the efforts of myself and a dynamic team of 5 software engineers. Working with a customer from the business department at Exeter University, our mission was to integrate the additional functionality of Econ Archive seamlessly into the Econ Plaza platform. Econ Archive introduced a dedicated section facilitating discussions on economic papers among users, with the added requirement that these papers and conversations be easily accessible from the main Econ Plaza page.

Throughout this endeavor, our team embraced an agile methodology, punctuated by daily stand-ups and weekly scrums. Ensuring continuous alignment with the project's objectives, we maintained bi-weekly meetings with the customer to gather feedback and confirm the project's ongoing alignment with the defined scope. In this collaborative environment, my role focused on the management of containerization and cloud deployment for the service.

To realize this, I employed tools such as Docker, Kubernetes, Google Kubernetes Engine, SQLite, and a Postgresql Database hosted in Google Cloud Storage. The success of our efforts was duly recognized, with the university awarding our group project a high first grade, reflecting the excellence achieved in the development of Econ Plaza's enhanced features.

Student Network

In the aftermath of the Covid Pandemic, a growing challenge faced many students, both new and returning, in connecting with each other at University. Our response to this need was the creation of "ReConnect," a sleek social network site designed to provide students and lecturers with a platform for interaction, infused with elements of gamification to encourage connections. Collaborating with a team of 5 other students, we successfully brought this vision to life.

ReConnect offered a diverse range of interaction options, allowing users to post thoughts, images, and videos on a feed visible to their conncections. The platform also featured the creation of quizzes, enabling users to gauge their own and their friends' knowledge. Adding a touch of fun, we implemented an achievement and level system that rewarded users for completing tasks centered around the theme of connection.

Adopting the Kanban methodology facilitated our project development, empowering us to focus on tasks that suited our strengths, resulting in the efficient creation of a high-quality social media site. The success of our project was highlighted by the university's invitation for us to deliver a talk on the design and implementation of ReConnect.

Alexa Clone

The objective of this project was to replicate the functionality of the Alexa voice assistant, achieved through the development of a series of microservices using GoLang. Each microservice played a specific role in the intricate system I created.

The first microservice was designed to listen for a request containing a base64 encoded audio file. Leveraging the Microsoft Speech to Text API, it converted the speech to text and responded with the extracted text. The second microservice, in turn, awaited a request containing a question, utilizing the Wolfram Alpha API to generate an answer, which was then returned as a response.

The third microservice focused on responding to a request containing an answer. By utilizing Microsoft's Text to Speech API, it converted the answer into a base64 encoded speech file. The final microservice was the orchestrator, listening for a request containing a base64 encoded audio file. It seamlessly integrated the other three microservices, essentially mimicking the functionality of the Alexa service, and returned the final base64 encoded speech file.

The use of microservices in this application offered notable advantages, allowing each component to function independently when necessary. This not only enhanced flexibility but also simplified the debugging process, providing a more streamlined and efficient development experience.