“HomeLight uses actual sales data and client reviews to find the best real estate agent for you”
I was the lead developer on the backend system driving the Homelight prototype. We worked with the new startup to turn their proven excel-based system of ranking real estate agents into a functional prototype that secured funding from multiple investors, including Google Ventures.
The core of the problem had two main components:
1) Collect data from multiple sources and create aggregate dimension tables suitable for use with the Homelight ranking algorithm.
2) Implement the algorithm in SQL to produce real results from the data.
This resulted in an exercise of iteratively adding to the dimensions we were aggregating then using them as part of a large Postgres CTE-based query until we reached a performance level suitable to take to investors.
The end product was a Ruby on Rails app that made heavy use of Resque jobs to handle the millions of sales records we were processing.
I’m really happy that it worked well and that Homelight have gone on to secure a further $3 million in funding to reach their full potential.