AI has shifted from a buzzword to a genuine, intelligent assistant. Host Daniel Kazani talks with Dajana Stojchevska, a senior software engineer at Google in Munich, about how AI is embedded in day-to-day engineering work, not about replacing engineers, but about boosting productivity and removing friction so teams can focus on the cool stuff.

ㅤ

Creation, collaboration, and knowledge management show up everywhere, from intelligent code completion inside the integrated development environments, to AI-assisted code reviews, to meeting notes that summarize transcripts, highlight key decisions, and list action items with owners. The conversation also remains grounded in the challenges: the hallucination trap, prompt injection, indirect injection hidden in external data such as a PDF, and strict discipline around data privacy. Looking ahead, the focus turns to autonomous agents, massive context windows, proactive analysis, and the evolving role of the software engineer as architect and orchestrator.

Dajana Stojchevska is a senior software engineer at Google in Munich. She graduated with a degree in Scopia from the Faculty of Computer Science and Engineering, with elective subjects in software engineering. She completed a few internships, including in Python, and her first role was as a Java developer focused on full-stack web development with Java and Angular. She also worked as a laboratory teaching assistant, helping students with exercises. After about two years, she moved to Germany for the Google offer.

ㅤ

1. AI inside the editor as a proactive teammate, intelligent code completion, prompts for snippets, and boilerplate
2. Code reviews with AI, drafting descriptions, style and standards fixes, and automated fixes from static analysis errors
3. AI-generated suggested code edits from teammate feedback, plus reviewer support with links to documentation
4. Meeting notes that summarize transcripts, highlight key decisions, and list action items with owners
5. Generating architecture diagrams from text, plus document analysis and “interviewing the document”
6. Correctness and the hallucination trap, treating AI like a junior engineer who needs supervision
7. Security risks, direct prompt injection, indirect injection, and why even a PDF can be a hostile input
8. Data privacy and strict guidelines on what data can go into which tools, plus internal AI chatbot support
9. Staying up to date with internal channels, newsletters, tech talks, hands-on daily practice, and peer community
10. The next five years: autonomous agents, bigger context windows, proactive help, and...