Sign In

Sergey Velts

Founder & CTO – Cybertonica

Sergey Velts is a highly accomplished expert in the fields of data science, computer security, and FinTech. He has a vast array of experience as both a founder and CTO of several successful startups, as well as a former Big Data Team Lead at Sberbank Labs at Sberbank.

Sergey's educational background includes attending the Yandex School of Data Analysis as well as completing his Ph.D. in computer security and machine learning with a 4.9 GPA at Московский Государственный Технический Университет им. Н.Э. Баумана (МГТУ).

Throughout his career, Sergey has worked as an Ассистент at МГТУ им. Н. Э. Баумана, a team lead and senior software developer at Diahelp, a software developer at МГТУ им. Н. Э. Баумана, a Java Developer at Jet Infosystems, and a System Analyst at БФТ.

Additionally, Sergey is the founder & CTO at Cybertonica Ltd, where he continues to be a driving force behind its success. He is a well-regarded name in the industry, and his skills and expertise are sought after globally.

Highlights

Jun 20 · linkedin

Nice article about BNPL in Middle East.

"The youngest demographic members prioritise personalised shopping experiences over concerns about data-sharing, in contrast to millennials who expressed greater reservations about sharing their data."

This is interesting. Personally, I'm definitely on privacy protection side. That's why I try hard to ensure that proper cryptography is present in all APIs, including (relatively) novel things such as oblivious transfer and PMT. Behaviour analysis is from the same category – then you have lots of PII you can probably do without behaviour analysis but try to do good scoring when you only know session info and phone number.

Just curious how much privacy GenZ is ready to give up for convenience.  For example, I know a bank, they are super convenient and innovative. One of their innovations is doing face recognition in mobile device’s phone gallery for scoring purposes. Maybe GenZ doesn’t not know (yet) what they are ready to give up...

https://lnkd.in/dhAhd28w

Apr 28 · linkedin

Session monitoring for better BNPL scoring

Credit scoring and fraud monitoring are related because different fraud attacks can result in defaults. That’s why we started to help BNPL companies using our session and behavior monitoring tech.

The idea is that we collect device fingerprints, behavior information, then do link analysis, feature engineering and then return Feature Vectors (dim > 100) to client’s scoring system in real-time using Score Session API. We also return indicators of compromise like bot, new device and so on.  The features give uplift because they help to reduce real-time fraud (etc. multiple applications by the same person from emulators, fraud rings and so on).

Finally, let’s look how neural networks use these features and try to separate good and bad BNPL loan application sessions. Below is a visualization of autoencoder after different number of epochs of training. Blue/orange - train, green/red - test good/default correspondingly.

#scoring #ml #bnpl #cybertonica #visualization

Related Questions

What is Cybertonica Ltd's main focus or product?
How did Sergey Velts start his career in tech?
What are some notable projects Sergey Velts has worked on?
How has Sergey Velts contributed to the tech industry?
What technologies does Cybertonica Ltd specialize in?
Sergey Velts
Sergey Velts, photo 1
Sergey Velts, photo 2
Add to my network

Location

Dubai, United Arab Emirates