Overview
Technical architect and engineering manager for CoinGenius from September to December 2019, covering the data platform behind their market analytics and the public website that presented it.
What CoinGenius was matters for reading the rest of this. They were a market reporter, not a token issuer — the model is closer to a financial data provider than to anything that creates or promotes a cryptocurrency. That distinction set the requirements: the product’s value was accuracy and timeliness of reporting on other people’s markets, which makes data quality the product rather than a supporting concern.
The platform ingested unbounded streams from sources that each behave differently — exchange market data, trading volume, social sentiment, news, and blockchain transactions — and normalized feeds that disagreed in format and timing into something queryable fast enough to serve real-time views and API endpoints. On top of that sat derived datasets, including indexes of cryptocurrency and token markets built in Python and R. An index is only as trustworthy as the warehouse underneath it, so a large share of the effort went into cleaning and consistency rather than into the index logic itself.
My core contribution was architectural and organizational. I containerized the Java, Python, and R applications with Docker and deployed them on AWS ECS, so every developer had one deployment pattern instead of three; added Apache Airflow to make the ETL steps explicit and schedulable; used Google Firestore as the platform’s database; and worked with the engineers across data sourcing, warehousing, ETL, front end, and data science to agree on the same tools and handling principles. The scaling pressure was not only data volume — it was more developers, more customers, and more tokens, and each of those breaks a system that has no shared convention. Alongside that I ran engineering management: sprint cadence, technical direction, and delivery.
The company ran out of funding. The team was capable, the thesis was sound, and the timing was early — and none of that survives a funding gap. It was a short, dense build rather than a long relationship, and the patterns worth reusing from it are about standardizing an environment before a team grows into its inconsistencies.