Integrations
Order, invoice and pricing data moving reliably between ERP, partners and customer-facing apps.
AI Engineering · Application & Integration Specialist
I build and operate reliable business systems — enterprise applications, EDI and API integrations, automation and observability — alongside practical AI engineering: structured LLM services, agent-accessible applications and production AI operations.
ilja@nedilko.net · nedilko.net · linkedin.com/in/iljanedilko · github.com/ilne
I’m a self-taught enterprise IT and integration specialist with 15+ years of hands-on experience across applications, automation, EDI, web development, and systems operations.
Experienced across the Microsoft and Azure ecosystem, with hands-on Linux, web-stack, database, and automation work. I focus on automation, observability, and reproducible operations — I treat environments like code, logs like truth, and outages like puzzles.
I apply AI where it delivers operational value rather than novelty: structured LLM services, agent-accessible systems, and LLM-assisted development connected to the business applications, ERPs and data platforms teams already depend on. What that looks like in practice is set out in the next section.
Order, invoice and pricing data moving reliably between ERP, partners and customer-facing apps.
Manual routines turned into repeatable pipelines that run unattended and leave an audit trail.
Applications that stay available, with incidents diagnosed and resolved before they escalate.
Agents given safe, metered access to real systems, so AI does useful work without new risk or runaway cost.
Practical AI engineering I have designed, built and operate myself — described as capabilities rather than products, because the engineering is what transfers.
Typed questions and calibrated probabilities · per-key rate and daily spend limits · compact responses · agent documentation generated from the implementation
Page-bound tools for signed-in applications · dry-run validation · idempotent writes · least-privilege boundaries that inherit the page’s own session
Server-side key custody · per-request spend ceilings · per-application cost attribution · output redaction and an explicit SSRF boundary
Browser audio capture, mono downmix and 16 kHz resampling · provider-aware options · store-before-send recovery and bounded retry
HTML, JSON and Markdown representations · idempotent writes · efficient polling · explicit human and agent handoff states
Versioned, reusable skills shared across agent harnesses · credentials kept outside source control
Worked on two parallel projects — Microsoft End User Fulfillment and ASUS:
Issued August 2013 · Credential ID E368-8211
Networking, Security, Windows OS, and Windows Server Administration Fundamentals. Issued August 2013.
Electrical Engineering and Automation, 2008 – 2009.
2004 – 2009.
1998 – 2004.