When a multinational accounting practice needed to transform how 3500+ clients produce financial statements under Bulgarian GAAP or IFRS, they didn't need another AI enthusiast. They needed someone who understood both the business domain and the engineering depth to build production systems around tech capabilities. That intersection — where deep operational knowledge meets modern software architecture — is where I've spent the last two decades arriving, and where I now do my best work.
I don't compete with AI. I build the infrastructure that makes AI operationally useful.
My career began in 2005 with banking internships at Raiffeisenbank and Societe Generale in Sofia, where I first encountered the gap between what organizations need from their data and what their systems actually deliver. By 2006 I had joined Boni Holding as a Marketing Expert, and within four years the role had evolved into something the original job description never anticipated: I became the person who automated the hard problems.
From 2010 to 2023 — thirteen years at Boni Holding — I built and maintained the company's core sales forecasting and analytical infrastructure. Over 15,000 lines of VBA drove forecasting models, automated reporting, and decision-support tools that executives relied on daily. These were not side projects; they were the systems the business ran on.
What those thirteen years taught me was not just how to write code, but how to encode operational knowledge into automated systems — how to translate the unwritten rules, edge cases, and domain expertise that live in people's heads into reliable, testable software. That skill would prove to be exactly what the AI era demands.
In 2020, I made a strategic decision to formalize and modernize my engineering capabilities. I completed an intensive program at SoftUni covering Python ORM, MS SQL, PostgreSQL, Java algorithms, and Linux system administration. This was not a career pivot born of necessity — it was a calculated investment. I had been building production systems for over fifteen years; now I was adding the formal computer science foundation and modern tooling to match the depth of my practical experience.
From November 2023 to May 2025, I worked as a System Software Developer for Databases and BI at DSK-Rodina, a pension insurance company. There I built a fraud detection system processing 500,000+ rows of transaction data using Levenshtein distance and graph theory. A sales force effectiveness analysis I created surfaced patterns significant enough to trigger an emergency board meeting and subsequent company restructuring. I automated PDF-to-database synchronization across 100,000 contracts, eliminating two days of manual work monthly.
Since November 2025, I have been building production software at Kreston BulMar, a member firm of Kreston Global, one of the world's top accounting networks. My primary deliverables are two interconnected systems: Mapping Studio and Reports Generator.
Mapping Studio is a PyQt6 desktop application that automates the mapping of client chart-of-accounts data to standardized BG GAAP financial statement line items. It implements 69 mapping rules, supports split mappings and debit/credit disaggregation, and includes a semantic auto-mapper that reduces manual classification work by an order of magnitude. The system serves 50+ active clients.
Reports Generator produces compliant financial statements — trial balances, balance sheets, income statements, cash flow statements, and general ledger reports — directly from mapped data. The validation layer enforces 17 BG GAAP validation rules, and the test suite exceeds 990+ tests covering the integration chain end-to-end.