Konstantin TarandevichPersonal blog and portfolio of Konstantin Tarandevich, Software Architect at ZF Group based in Koblenz, Germany.https://site.tarandevich.org/Everyone Is Building an AI AUTOSAR Toolchain. But How Should We Evaluate One?https://site.tarandevich.org/blog/everyone-is-building-an-ai-autosar-toolchain/https://site.tarandevich.org/blog/everyone-is-building-an-ai-autosar-toolchain/Generating plausible AUTOSAR artifacts is an interesting capability. It is not yet evidence of a production-ready engineering system. We need a way to measure the difference.Sun, 19 Jul 2026 00:00:00 GMTHybrid Databases Were the Wrong Fix for Requirements RAGhttps://site.tarandevich.org/blog/hybrid-databases-wrong-fix-requirements-rag/https://site.tarandevich.org/blog/hybrid-databases-wrong-fix-requirements-rag/A negative infrastructure result from a requirements RAG lab: Postgres, ParadeDB, and OpenSearch could host useful pieces, but none replaced the current retrieval stack without losing ranking quality or evidence bundles.Wed, 24 Jun 2026 00:00:00 GMTThe RAG Verifier Learned to Say No. It Still Missed a Hard Requirement Failure.https://site.tarandevich.org/blog/claim-level-verification-requirements-rag/https://site.tarandevich.org/blog/claim-level-verification-requirements-rag/A held-out requirements RAG experiment: claim-level verification reduced harmful published answers from 35.7% selective risk to 4.0%, but strict human audit still found semantic completeness failures.Sun, 21 Jun 2026 00:00:00 GMTCitations Were Not Enough for Safety-Critical Requirements RAGhttps://site.tarandevich.org/blog/citations-were-not-enough-safety-critical-requirements-rag/https://site.tarandevich.org/blog/citations-were-not-enough-safety-critical-requirements-rag/A negative result from a requirements RAG lab: even with structured records, cited JSON answers, and deterministic guards, the answerer failed the abstention gate when evidence was insufficient.Mon, 15 Jun 2026 00:00:00 GMTI Flattened 13,244 Requirements Into Chunks. 97% Lost Their Meaning.https://site.tarandevich.org/blog/rag-structured-requirements-router/https://site.tarandevich.org/blog/rag-structured-requirements-router/A practical retrieval study over structured automotive requirements: why flatten-and-chunk failed, where ordinary top-k search stopped being the right tool, and how query routing improved both quality and latency.Sun, 07 Jun 2026 00:00:00 GMTI Tested a Simple RAG Idea: One Summary per Document. It Was Useful, But Not Enough.https://site.tarandevich.org/blog/rag-retrieval-lab-summary-retrieval/https://site.tarandevich.org/blog/rag-retrieval-lab-summary-retrieval/A practical retrieval experiment: can an LLM-generated document summary replace chunk-level search? On a 5,000-document corpus, the answer was measurable — useful signal, clear ceiling, and a better baseline.Sun, 31 May 2026 00:00:00 GMTMy Two AI Agents Talk MCP to Each Other. There Is No Standalone Tool Server.https://site.tarandevich.org/blog/agents-talking-mcp-without-a-tool-server/https://site.tarandevich.org/blog/agents-talking-mcp-without-a-tool-server/The default MCP story is one fat tool server, many clients. When I needed two of my domain agents to talk to each other, that shape was the wrong one — and the alternative shows what MCP is actually good for.Thu, 28 May 2026 00:00:00 GMTI Built an AI Agent for Myself. My Colleagues Wanted to Use It. That Is Where the Problems Started.https://site.tarandevich.org/blog/ai-agent-bus-factor-of-one/https://site.tarandevich.org/blog/ai-agent-bus-factor-of-one/When a domain expert builds a working AI agent and people start depending on it, the bus factor problem arrives long before the platform — and a different kind of work begins.Thu, 14 May 2026 00:00:00 GMTThe Boring Part of Requirements Review Is Now Automatedhttps://site.tarandevich.org/blog/ai-agent-requirements-review/https://site.tarandevich.org/blog/ai-agent-requirements-review/How a lightweight AI agent eliminates the context-gathering overhead in safety-critical requirements review — so the architect can focus on judgment, not tool-switching.Sat, 02 May 2026 00:00:00 GMTMeasuring What Matters: Architecture Quality Metrics for Safety-Critical Softwarehttps://site.tarandevich.org/blog/measuring-what-matters-architecture-quality-metrics/https://site.tarandevich.org/blog/measuring-what-matters-architecture-quality-metrics/Architectural quality is difficult to observe — you cannot run a test suite on a dependency graph. This post describes a metrics system built to track it: three levels of measurement, from binary compliance checks to long-term trends.Tue, 28 Apr 2026 00:00:00 GMTBuilding an AI Agent for Embedded Systems Architectshttps://site.tarandevich.org/blog/building-ai-agent-for-embedded-systems-architects/https://site.tarandevich.org/blog/building-ai-agent-for-embedded-systems-architects/A conversational AI agent that sits in front of complex embedded systems tooling and speaks human — so architects spend less time remembering plugin syntax and more time on actual architecture.Mon, 27 Apr 2026 00:00:00 GMT