Pavlin Petkov
Senior Software Engineer at SofiaStars
Berlin, Germany
Actions
I’m a Senior Software Engineer with over 10 years of experience designing and building scalable backend and full-stack systems, primarily in fintech, payments, and large-scale platforms.
My work focuses on payment infrastructure, event-driven architectures, and reliability-critical services, including the design of payment gateways, transaction flows, and integrations with multiple third-party providers.
I’ve worked across startups and large international companies, including five years at Wayfair in Berlin and more recently at SofiaStars, contributing to systems that operate at scale and serve millions of users.
My core expertise lies in building production-grade systems, with a strong emphasis on scalability, resilience, and clean architecture.
Area of Expertise
Driving AI out of Business by using AI
AI made a lot of people get comfortable with generating code, but a lot of them, especially new developers, might not know that we used to have auto generating tools without AI as well - namely JHipster in the Java world and Maker Bundle in PHP.
They could generate you a whole CRUD application, generate Entities with ORMS, create migrations, service layer, validators, controllers with documentation. Even basic html templates for filling out forms and viewing data. All without AI.
But most people didn't use them at all...mainly because they relied on strict architecture and your codebase had to follow the best practices in the respective frameworks - Spring & Symfony respectively.
You could configure them, but if you worked on a large project with years of legacy code that was very difficult to do and most developers didn't use them or never knew about them.
Currently AI costs are skyrocketing, free plans are being demolished and become more and more limited in their functionality.
So what if we used AI to build us more flexible versions of JHipster and MakerBundle? We could generate workflows with command line interface that are specific for our legacy enterprise applications.
That would still be very rigit. Commands with 100 options and command parameters. So what if we had a local AI model which handles the communication with our command line commands?
The benefit? Why bother at all?
- Security! Takes care of prompt injection and similar issues.
- Much less AI usage. Once terminal commands are defined we no longer use paid large LLMS.
- Code Quality - we define templates and requirements and the workflows are more concrete. AI can pass different parameters to predefined workflow so the actual code is defined by you - in the templates, which AI uses to build the commands.
- Significantly cheaper - as prices for LLMs keep skyrocketing and free plans offer less and less functionality this approach can save us a lot of money.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top