Vector Space Signal

16. May 2022.

Built with: Next.js, d3.js, Tailwind CSS, Django

Categories: Computer Networks, Web Development

Vector Space Biosciences is an NLP and academic data science company operating in the space biosciences industry. They use patented NLP/NLU technology to discover hidden relationships using real-time data pipelines.

Vector Space Signal

Vector Space Signal is a data-driven web application using near real-time data engineeering and NLP to display hidden relationships based on how global events will impact a portfolio, generate thematic short baskets of stocks and cryptos, and predict convergence between stocks. The demo application was invited to be presented at Morningstar Investment Conference 2022.

Outcome/Role

I had the role of data visualization and product engineer and was pivotal to the end-to-end product development where I was involved from the design, development, and deployment of the product.

Product Design

As a designer, I began the project from initial concept where I conducted user interviews with financial analysts and experts to discover how quantitative analysts typically use trading software tools in their line of work. From these interviews I created user stories of different categories of users, from retail investors to M&A financial advisors, to develop product features to highlight how NLP and machine learning can be used in the financial industry. From these user stories I created lo-fi and middle-fidelity wire frames to communicate the data requirements to the NLP Engineering team and Data Engineering team.

Product Engineer

As a product engineer, I worked closely with a data engineer to implement the web application, where my primary responsibility was the front-end development of the application. In this role, I coordinated with a data engineer to outline API requirements to drive the front end of the application where the APIs were built with FastAPI. I developed the front-end interface using Next.js, TypeScript, and Tailwind CSS. I used D3.js to parse data retrieved from the API to create hierarchal treemaps and reusable line graphs.

Deployment

As the primary web engineer I was responsible for developing the deployment system. I provisioned a Cloud Compute Engine VM on GCP, and created a multi-stage build container network using Docker. I created separate containers for the FastAPI service, web app, and an NGINX service which was used to proxy API requests to the FastAPI service. Finally, I implemented a Certbot container for SSL certificate for the web application.

Authentication

After the conference, I worked to further productize the application by creating a user account system for the application where I contributed as a full stack developer and designer. I design a login and sign up flow and built an authentication system using Django-Admin and Django Rest Framework to pass JWTs to the Next.js web app using React Context API to manage authentication state.