töökratt
I wanted an assistant that reads the current job market and shows which skills are actually in demand, so people can re-skill against real postings instead of generic career advice.
I built a RAG pipeline: ingest listings from The Hub API, embed them in Qdrant, and ask questions over the index (FastAPI, Gemini 2.5 Flash) with citations. The first production request OOMed because the embedding model loaded in-process on a 512MB box. I moved embeddings to Qdrant Cloud Inference, benchmarked two free models on the real 1,015-document corpus, rejected one because its scores couldn't support a hard cutoff, and reindexed without taking the assistant down.
The assistant is live. I also run evals (unit, retrieval, generation) and Grafana/Loki, including an alert on prompt-injection attempts.