Systematic_me: an optimal systematic experience (SX) that springs new digital opportunities

Sergio Alvarez-Teleña is Head of Global Strategies & Data Science at BBVA and CEO at inSCIghts, and a former student at the EPSRC Centre for Doctoral Training in Financial Computing at UCL.

In a data-driven world, being systematic is crucial. As such, machines are smoothly taking over human responsibilities. They are progressively reaching full autonomy on key human decisions that often imply legal representation of corporates at high speed (e.g. eCommerce). And machines can have thousands of possible configurations that we can expect will allow it to execute those tasks "correctly" (offline, based on backtested, controlled scenarios). The question is: how can the user select the configuration that optimises their online experience?

Avatar learning translates the outcome of each combination of parameters within the calibration universe into a "robot's brain" that is then compared with the user's brain. The closest brain is hence selected and its combination of parameters used in production.

A few years ago I left the industry seeking to formally address this puzzle. I undertook a PhD in Computer Science at the EPSRC Centre for Doctoral Training in Financial Computing at UCL and started researching on a field that boasts of being at the forefront of technology and innovation: Financial Computing. And the methodology that I finally authored, Avatar Learning, allowed a trader (the user) to discover a combination of parameters of a high frequency strategy (the speed) that naturally led to a prudent pattern exploitation policy (blending data-driven and expert-driven techniques); and ultimately enhanced the robustness of its performance online (the experience). This improvement was based on the search for a robot that best mimicked the trader's brain (a hypercube of core check points) through Reinforcement Learning. More interestingly, the technique is flexible enough to be applied onto several other domains.

In fact, I am now experimenting on its real-world implementation within two corporates. First, BBVA. BBVA is a reference in banking digitization. Its management team believed in my thesis to improve their trading approach within Equities Europe. As I started delivering I was gradually promoted onto a global and cross-asset role with responsibility upon both smart trading and eCommerce - quite a unique position in the industry. My team includes highly skilled developers and a former senior lecturer in Machine Learning. The results so far are remarkable. Publically I can only say that we can reach bid-offer spreads reductions of up to 50% and we can save the cost of development of the strategies by multiples of ten. The plan is to keep growing my team's responsibility across the data-driven businesses. Second, inSCIghts - my startup. It is the core hub where I can further research on how to exploit Avatar Learning (we call it "Systematic_me") on key domains such is cybersecurity - actually, inSCIghts is currently helping innovate this area with BBVA as client.

The mid-run target is to let retail inherit this technique and take it one step forward. We plan to massively help individuals take more educated systematic decisions on fields where they may not feel comfortable (wealth management, daily expenses, insurance, security...) by letting them mimic the essence of the smartest individuals. For example, they could simply rent the Systematic_me of the most skilled individuals on those fields. It is hence a greenfield for business innovation that could definitely bring a completely new source of digital revenues.