Digital Twin Decision Maker (DTDM)

DTDM builds your decision‑making twin from your preferences, then uses agents to recommend what you'd do with clear reasoning, confidence, and regret—transparent, demo‑ready AI.

Comment

Inspiration

We wanted to make AI advice feel personal and trustworthy, not generic. The inspiration came from seeing people make major life and product decisions without a consistent way to capture their own values, context, and past choices. DTDM turns that intuition into a digital twin that can reason like the user and make the decision trail auditable.

What it does

  • Builds a personal decision profile from preferences and onboarding answers.
  • Classifies each decision and retrieves similar past choices.
  • Generates a first‑person recommendation plus explanation.
  • Returns confidence and regret scores for risk-aware decisions.
  • Stores and surfaces a history of decisions.

How we built it

  • Backend: FastAPI with agent orchestration services, modular routers, and mock mode for reliable demos.
  • Agents: Decision flow that chains classification, memory retrieval, reasoning, and scoring.
  • Data layer: Firestore model for profiles and decisions, with mock storage for local testing.
  • Frontend: React + Vite with a demo-first UI, scenario presets, and theme toggles.
  • Infra: Terraform baseline for Cloud Run, Firestore, and Secret Manager.

Challenges we ran into

  • Balancing a compelling demo experience with realistic end-to-end architecture.
  • Designing a decision flow that feels personal without real user data.
  • Making the UI polished while keeping the data flow simple and stable.

Accomplishments that we're proud of

  • A full-stack MVP that demonstrates agent orchestration end-to-end.
  • A demo UI that is stage-ready with scenarios, theming, and clear outputs.
  • Clean separation between mock and production-ready integrations.

What we learned

  • Clear UX and storytelling are as important as model outputs for trust.
  • Agent pipelines need explicit contracts to stay maintainable.
  • Building for demo reliability early saves time in final presentation.

What's next for Digital Twin Decision Maker (DTDM)

  • Wire real Firebase OAuth providers in production
  • Replace mock agent calls with Vertex AI endpoints
  • Add feedback learning and outcome tracking
  • Add mobile‑ready UI

Built With

Share this project:

Updates