Thomas Manley

I designed, developed, and deployed a full-stack web application for an influencer marketing platform. The front-end app was built using React and Redux, while the back-end app with Ruby on Rails. I used MySQL for relational data and Redis for storing non-persistent caching and queuing information. The application was deployed to AWS and successfully supported tens of thousands of users.

I built a RESTful API to perform >99% accurate email mailbox validation. The application accepted an email address as its only argument, retrieved MX records from DNS servers, opened a Telnet connection with the appropriate mail server, and queried the user's mailbox. Status codes were returned for common faults, including "address does not exist," "mailbox full," "honeypot," and"spam trap," to name a few. The API was used to drastically improve email deliverability for Hola Influencer Marketing and completely in-sourced the ZeroBounce business model that was previously utilized. The API was deployed as a serverless script on AWS Lambda.

I developed an API that performed image recognition using TensorFlow to accurately identify whether a given image was inappropriate (not safe for work). The API accepted an image URL as its only argument and returned a JSON response with five categorized predictions. The API was deployed as a serverless script on AWS Lambda. It was used by Hola Influencer Marketing to successfully hide or blur images that contained nudity (full or partial) or sexually explicit content.

I created a messaging interface using React and Redux that received and displayed real-time message notifications from Socket.IO. The interface was used by Hola Influencer Marketing and allowed brands and influencers to collaborate on sponsored content in real-time. The messaging portal allowed users to send text, embed images and videos, links to external sites, and inject contractual events such as payouts and deadlines.

I built a search system in React that used virtualized windowing to limit the size of the DOM tree (for performance) and infinite scrolling. Virtualized windowing allows you to store hundreds or thousands of DOM elements in a cache while only rendering the ones that should appear in the viewport. This allowed for drastic improvements in scrolling consistency and image rendering. The infinite scrolling improved user experience and increased profile clickthroughs by 34% versus traditional pagination.