The Nokia ML system bullet (“multi-agent LLM system…”) sounds cool but is kinda vague on what actually improved. I’d explicitly say what changed (e.g. faster resolution time, % reduction in incidents) so it doesn’t read like a research project.
The dashboard bullet (“cutting report generation time from 10–15 min to under 1 min”) is good, but it’s unclear what you actually built to achieve that. I’d briefly mention the key change (e.g. caching, query optimization, async jobs) so it shows real engineering depth instead of just the result.