We are building a better way to get a mortgage. Fully online, with no commissioned sales people. We fetch data from APIs instead of asking users to upload a million docs. We evaluate 1000s of rules automatically to match users with capital. Using automation we can lend money in a fraction of time and with a fraction of the cost of a traditional lender. The end result is lower rate, better experience, and a much faster process.
Why join us?
We have raised $45M from investors like Kleiner Perkins and Goldman Sachs
You might not have thought much about the mortgage industry, but it's one of the largest industries in the US, and massively broken. 7 million Americans get a mortgage every year, it takes 8 weeks on average, and the cost is $10k
We have built an amazing tech team led by the former manager of the machine learning team at Spotify
Engineering at Better
Since day 1 we've used continuous deployments and we deploy to production 40-50 times every day. When an engineer is done with a task, they create a pull request, which kicks off a extensive test suite. Our own system assigns a reviewer to the the pull request. Assuming the pull request is approved (usually within 15-20 min) it's merged into master and goes live into production in another 15 minutes.
Everyone here is a full stack engineer although most people skew very heavily towards backend. The tech team is split up into three smaller feature focused teams that do daily stand ups but other than that we have very few meetings or other process overhead.
At the core of our system is a very complex rules system that encodes 1000s of rules and figures out the next steps in the process. We integrate with probably 70 different APIs for things like bank statements, tax returns, property data, ordering appraisals, and much more, so at any point we need to coordinate everything and deal with failures and retries automatically.
Mortgage origination is fundamentally an objective decision, but for historical reasons it's always been done by humans. We have an automated underwriting system, which is actually turns out to be an NP-complete problem, but you can rewrite it to a mixed integer programming problem that runs very fast.
We are increasingly starting to look into technologies like OCR for parsing structured documents.
This is one of the biggest financial transactions most people make in their lifetimes. To help people understand the choice they are making, we have a bunch of complex D3/React widgets that visualize various financial decisions.
An ongoing project right now is to automate underwriting. There are several different steps of it, but roughly 1. Get the data from the customer. We try to use APIs whenever we can, but still much of the data is in the form of PDF. We are planning to use OCR to read data from documents (paystubs, tax returns, etc). 2. Compute a number of derived metrics. Something like
incometurns out to have a very complex definition governed by 1000s of rules. Same with assets, debt payments, and other metrics. 3. Make a decision: what is the max this borrower can afford? If the borrower is not eligible, can we recommend ways to become eligible? These decisions turns into a nontrivial search problem in a high dimensional space.
Building our v2 Frontend: We started out on Ember and it has taken us quite far but since it's a single page application, it was not the best choice for the start page due to the slow loading time and poor SEO. We have factored out all of the static pages into React that we render server side in a separate container. We are also building our own internal stylekit (think of it as a Bootstrap version) that we have used for the React pages that we are now integrating back into the Ember app: http://stylekit.bettermg.com/
Working at Better
We iterate very quickly – it's not uncommon that we launch 5 or 10 major features in the same week. This is a performance culture where people join us because the want to work with other really smart engineers and learn things from each other.
We offer free lunches and dinners.
If engineers are productive, they can work when they want and where they want.
Work from Home
As long as people get things done, they can work anywhere they want.
Interested in this company?
Skip straight to final-round interviews by applying through Triplebyte.