My route into AI did not begin with a model
My professional path began in technical support, factory infrastructure, production planning, and textile manufacturing systems. Long before founding Aventra Labs, I was learning how software succeeds or fails when it meets real operations: orders arrive late, production capacity changes, materials move between facilities, and users need answers while the system is under pressure.
That background continues to shape how I approach AI products. A useful system is not defined by the novelty of its technology. It is defined by whether data is dependable, workflows reflect reality, decisions remain explainable, and the product helps people complete meaningful work.
Fifteen years inside textile ERP/MRP
At Egemen Software and Automation, I combined product ownership, product management, project management, business analysis, software development, and database architecture. I worked across implementations serving 36 integrated or independent factories and provided technical or operational consultancy to more than 100 companies.
Textile manufacturing is not one uniform workflow. Spinning, weaving, knitting, yarn dyeing, package dyeing, fabric dyeing, finishing, printing, laboratory work, planning, costing, procurement, and dispatch each create different constraints. Building an ERP/MRP product required translating those constraints into a shared data model without erasing the operational differences that mattered.
Product ownership meant understanding the factory floor
Requirements were rarely complete when first expressed. A request for a report could reveal a missing production event; a planning problem could originate in master data; a costing discrepancy could expose an integration gap between production and accounting. My role was to move between executives, planners, laboratory teams, production operators, and developers until the underlying problem became clear.
This taught me to treat business analysis as system design. A backlog item is not valuable because it is detailed; it is valuable when it represents the real decision, dependency, or exception the product must support.
Public-sector analytics added another scale of complexity
Between 2020 and 2025, I worked as a senior full-stack software engineer on an EU-funded workforce and workload analytics platform for the Ministry of Labour and Social Security. The project moved from greenfield architecture and process digitalization to dashboards, performance optimization, security hardening, user acceptance testing, and final delivery.
Working with C#, Blazor, PostgreSQL, PL/pgSQL, Syncfusion, and DevExpress strengthened my experience in analytical applications where many operational processes must become reliable metrics. The platform's recognition through two public-sector digital transformation awards reinforced a lesson I had already learned in manufacturing: good analytics begins with good process modeling.
How those lessons became Aventra Labs
Aventra Labs applies the same discipline to financial-market research. A chart pattern scanner may look very different from a textile ERP system, but the product questions are familiar. What is the authoritative entity? Which events are trustworthy? How should exceptions be handled? What should be computed once, and what must remain interactive? How can the user understand why a result exists?
AI expands what the product can help a user explore, but the foundation remains product ownership, domain modeling, data engineering, software architecture, and continuous feedback. My career has moved across industries, yet the central work has stayed consistent: turning complex systems into useful, explainable products.