TasaractDavut Emre Taşar
English

Profile

I’m a data scientist on BBVA’s AI Innovation team in Madrid. For my PhD at the University of Navarra, I’m studying whether measuring and reducing uncertainty in banking models can help people make fairer decisions.

View raw data JSON

About me

I’m from İzmir, on Turkey’s Aegean coast. I used to be a civil engineer. Then I took an arrow to the knee and traded my hard hat for a lab coat. These days I’m a data scientist on BBVA’s AI Innovation team in Madrid.

A bank can act on a model’s prediction even when the model is unsure. For my PhD at the University of Navarra, I’m studying whether measuring and reducing that uncertainty can help make those decisions fairer. Studying strategic management at Oxford got me thinking about the person making the call. Will a decision that looks sound today still hold up if circumstances change?

One thing I love about abstract maths is finding the same idea underneath two very different problems. Quantum physics tends to send me off into philosophy in much the same way. Start asking what a measurement can actually tell you and it’s a surprisingly short trip. I like finding those connections.

I’m still getting to know Madrid. Learning Spanish lets me discover a little more of the city, and of the cultures of Spain and Latin America. I also like painting and playing chess.

I make a lot of coffee at home. I have a small darbuka, a kalimba and, as a Bleach fan, a Zanpakuto that I use for kendo practice. Whenever I get the chance, I make up stories to make my daughter laugh.

Oxford Strategic Management Executive Programme

Work and study

BBVA · AI Transformation · AI Innovation
Applied Engineer / Scientist
University of Navarra
Doctoral researcher in AI fairness

Interests

Ontologies and data structures

What changes in a graph when a train runs late?

A station is a place. A departure is an event. If I give each its own node type and each time a source, can I trace exactly which connections a single delay affects?

Imagine two trains due to leave the same station at nine. In a graph I can represent the station once and give each departure its own identity. That lets me attach a revised time to one train without changing the other.

I would keep the scheduled time, the update and the source of each. If a departure moves from 09:00 to 09:10, I can follow the links from that event to see which onward connections need checking. A single timestamp on the station would hide that distinction.

RDF 1.1 · W3C
Meaning-Native Computing

Can a computer predict which instructions will run next if it knows the rules of the program?

Meaning-Native Computing is my independent research project. I want to compare two predictors given the same program and comparable computing resources. One also gets an explicit description of the rules the program must follow. Does that help it predict the next steps? Before a predicted change can take effect, an independent check must confirm that it follows those rules.

For my independent project, I want to give two predictors the same view of a running program and comparable computing resources. I give one of them an executable ontology, a description of the program’s rules in a form the computer can apply. Does knowing those rules help it predict which instructions will run next?

A program might allow two different next steps. A predictor could pick the wrong one and still obey every rule. I want to compare its prediction with what the program actually does. An independent check must establish that any proposed change is allowed before it can take effect.

Meaning-Native Computing
Category theory

What survives when I compose two transformations?

One process turns a timetable into a graph. Another finds a route through it. When I compose them, how do I preserve which station each time belongs to?

One process turns a timetable into a graph. Another finds a route through that graph. Putting the two together is an example of composition. What I want to know is whether I can still trace each time on the final route back to the right station.

Suppose two stations both list a departure at 09:00. If the first process keeps the time and drops the station, the route finder cannot recover that connection from the time alone. Even this small example makes me think about what each step needs to carry forward.

Seven Sketches in Compositionality
Quantum measurement

What if I ask the same qubit a different question?

I can prepare a qubit so it always gives the same answer when measured along one axis. Along a perpendicular axis, either of two results is equally likely. What fascinates me is how changing the measurement changes the answer while the preparation stays the same.

Imagine preparing a qubit in the plus state afresh before every measurement. In the X basis, an ideal measurement always gives plus. In the Z basis, the same preparation gives 0 and 1 with equal probability.

Ten Z measurements could give seven zeros and three ones. The chance of a zero is one half each time. That still leaves room for many different counts in a batch of ten, something that is easy to overlook in a short run of results.

Quantum Information · IBM · Specify observables in the Pauli basis · IBM Quantum Documentation

An award for my work

I received BBVA’s We Think Big award in 2024 for my work on patents and my academic contributions.

See the award photo

Contact & profiles