Alex Brad

An automated system for gathering, processing, and exploring cryptocurrency-related data. The data was gathered from various sources, text data from social groups, trading data from APIs, and data extracted from the blockchain.

Built a React web app for data exploration and visualization involving interactive charts and drill-down workflows. I was the sole developer for the project and led the research, design, and implementation of the end-to-end solution.

The drive behind the project was to be able to gather publicly available data that might indicate the economic activity of various public and private companies.

This involved a lot of research to identify potential indicators and then building automated scrapers to gather that data regularly. This would sometimes involve a lot of complexity because the workflows required to generate data were not straightforward.

The architectural focus of the system was to be very modular, so adding or replacing new data sources would be as straightforward as possible. Additionally, the system included a pipeline for transforming, validating, and aggregating the gathered data.

Initially, I managed the design and building of the system from scratch. Later on, when two additional people were onboarded, I was responsible for defining the research and data-gathering workflow and supervising the work of the two additional team members.

At Migacore, we've built a product that can help predict the impact of events on Airline demand and price sensitivity based on data mined from the web.

In short, our approach involved crawling structured and unstructured data to extract multiple data points related to events. These were later used to understand which events impact travel and what that impact is likely concerning demand and price sensitivity.

Migacore is a small startup, so I was involved in most areas of research and implementation, including:
• Building several scrapers to collect structured and unstructured data from the web.
• Working on the data pipeline that ingested this data, used Named Entity Extraction for isolating events, and assigned additional data points to events.
• Designing and training models to filter down many events to a subset of potentially relevant ones.
• Processing, exploring, and cleaning-up airline data.
• Designing and training models to predict the actual impact of the events on demand and price sensitivity.
• Mentoring other developers.
• Collaborating in building a web app and deploying our models and pipeline in a Kubernetes cluster.

Recognos had an in-house project for Semantic Travel Blogging. This involved using named entity recognition and sentiment analysis to identify and rank accommodations, restaurants, attractions, and activities.

I was in charge of the architecture for the web app and was involved only sparingly in the back-end NLP work. In addition to acting as a usual blogging platform, the app also had to provide tools for manually annotating text, which I built from scratch.

I also created a WYSIWYG web text editor inspired by the one Medium had at the time because none of the existing open-source ones satisfied our requirements.