What if a Bank was Run by Machines?

In a world where automation and machine intelligence continue to advance at an exponential pace, we believe the financial sector is ripe for reinvention. Can we imagine a bank that is completely operated by machines? This is our vision, our "utopia", and it's what drives everything we do. Below, we'll explore both the vision and the highly technical challenges we're tackling to get there, and why engineers find this journey so exciting.

Why should machines run a bank?

All the inefficiencies of our current banking system lead to costs - costs that are in turn passed on to users like you & me in the form of interest rates & fees. We believe that operating a bank autonomously - by moving electrons instead of moving physical atoms - will be the most efficient. This efficiency will lead to the lowest cost of capital for consumers. Running 24/7, 100% transparent, always auditable, observable, debuggable, trace-able & programmed for compliance, will make it the lowest cost to consumers both immediately and in the long term.

The Engineering Challenges

The Engineering Challenges of building this are enormous. They range from Machine Learning / Artificial Intelligence to Robotic Arms as we seek to find the absolute limits of what is possible. Below is a sneak preview of just a few of the types of problems we work on at Aven:

i. ML at Scale

Capital is a commodity, and in a commodity market, even a 1% efficiency gain is huge. Imagine an auto loan that is 1% cheaper - 90% of users in the world would switch to that lender. This means that using automation & machine learning maximally is the key to becoming a true machine bank. At Aven, Machine Learning isn't just a 'nice to have', it is at the core of what we do and seek to achieve. It is the backbone of our efficiency strategy. We spend millions of dollars annually on marketing and growth. Our efforts here involve using a sophisticated, multi-model ranking of the US population every 7 days. The ranking aims to predict each individual's propensity to respond to our messaging, which we then use to create highly personalized creatives that highlight our value proposition for each user. The impact of our modeling & sciences team work here saves us millions of dollars, and grows the balances we generate revenue from by billions.

ii. Robotics

Mortgage & auto loans in the US often require wet-signatures (completed with pen & ink). We invented robotic arms to help users remotely sign documents, a solution that involves real-time streaming of your signature to robotic arms. Blending robotics, legal research, modern UX-design, encryption, and distributed systems design - we invented our way around the inefficiencies of a legacy industry - unlocking technology that is now accepted by counties nationwide.

iii. Capital "Logistics"

We manage billions of dollars of credit lines for our users - that need to be allocated across a multitude of "debt" warehouses that are each in the hundreds of millions of dollars. How do you do pack these debt lines efficiently? This becomes a sophisticated constraint-solver problem - where each warehouse has different prices, and different limitations on who/what/how much it can serve. Optimizing saves our users millions, and can make the business millions of dollars. From random seeding, to simulations, to automatic projections of future user-acquisitions - this is fascinating problem to iterate on and make more efficient.

iv. Sentinel Runner Agents

Automated regulatory compliance is critical to our systems as we scale. Similar to NASA's space shuttle program, we run completely multiple, independent production grade systems to ensure that our core production system is compliant with every single state & federal regulation. Our independent agent code scans every user interaction every 4 hours to ensure that our production systems are compliant - from checking the delivery of any messages we sent to cross-checking the content of every single account agreements that people have signed with us.

v. Humans as Sensors

"A System of Machine-Orchestrated Work Queues" - is what we call operations. At Aven, we've introduced a model where machines coordinate nearly every operation, and humans act as sensors. All income tasks are funneled by machines into the appropriate task queue (called QUs). Humans are leveraged as sensors - ie detect fraudulent identity (ie man shows up wearing a hat and fake mustache - really!) and feed data into our machine systems which centralizes coordination and final decision making. This "humans as sensors" strategy strikes the balance between automation and adaptability. Machines handle day-to-day operations with minimal error and process massive amounts of data at scale, while humans step in when things get fuzzy.

Conclusion

For engineers seeking to push the boundaries of what's possible in Machine Learning/AI, Software engineering and financial technology - and want to work with incredible sharp, mission-driven, hard-working people - Aven is an opportunity of the lifetime.

At Aven, you'll be surrounded by extraordinary talent—people who have:

1. Built & led the first Applied Computer Vision teams at Facebook
2. Built fastest C++ compilers at Microsoft
3. Were the founding engineers of Azure at Microsoft
4. Were the founding designers on Cash-App at Square

If you're an engineer with a passion for solving impossible problems and want to work with a team that's rethinking the very nature of finance, we're building the future—join us.