Ishita Daga is a senior machine learning engineer at Tesla developing enterprise AI agents that can distinguish reliable business knowledge from outdated information and conflicting metric definitions. She treats organizational context as infrastructure: an agent cannot answer accurately without knowing which information to trust, whether it remains current, and whose interpretation applies.
Previously an AI scientist at Covera Health, Daga worked on machine learning for radiology and coauthored research on extracting diagnoses from cervical-spine reports. Her writing on few-shot and meta-learning explored how models learn when reliable labels are scarce.
Her approach to enterprise-agent architecture concentrates on three structural challenges:
- Rank knowledge by reliability. Start with curated semantic layers containing approved business definitions and calculations, then use canonical queries and, where additional flexibility warrants the maintenance, database graphs.
- Keep context alive. Connect agents to actively maintained systems such as GitHub, Tableau, and dbt; capture corrections; and test responses against human-reviewed or automated evaluations.
- Resolve team-specific metric preferences. Different teams can measure the same milestone between different events and obtain equally valid answers. Semantic layers and agent memory preserve alternatives, but neither reliably determines which definition a particular person intends.
Daga also created Savera, a solar-planning application combining rooftop analysis, California installation pricing, utility estimates, and battery-storage calculations. She described building it with AI-assisted coding and taking the project from concept to production in two weeks.
