Agentic LLM Testcase Triage Framework
Automating IVI/ADAS failure triage with retrieval-augmented agents and vector search
During my 2025 internship with Qualcomm’s Automotive IVI/ADAS team, I designed an agentic retrieval-augmented generation (RAG) workflow that triages more than ten thousand integration tests per day.
Highlights
- Consolidated multi-source IVI/ADAS logs into concise error narratives using large-language-model summarization seeded with error-anchor context windows.
- Indexed structured traces inside ChromaDB/FAISS to enable hybrid semantic + keyword retrieval that surfaces the most relevant comparator incidents.
- Coordinated LangChain agents, function-calling tools, and JSON-schema emitters so downstream dashboards receive machine-actionable diagnostics within minutes of test completion.