Goal: tips are AI-generated from user context, not just raw Todoist tasks. Multiple signal sources feed a generalized pipeline. Research-intensive milestone.
Key themes:
- Signal source abstraction (beyond Todoist)
- AI tip generation via local LLM (Ollama)
- Generalized candidate → rank → render pipeline
- Feature registry + profile builder
- Policy research (Thompson sampling, neural bandits)
- Reward model improvements
iOS + Android apps, push delivery, quiet-hours notifier.
Orchestrated retrains, MLflow registry, A/B + bandits, cohort features, drift monitoring.
Audit, token rotation, k8s + autoscaling, multi-region, public integration SDK, billing.