Jayda - Personal Terminal

Jayda OSS

Jayda — Personal Terminal for Care Navigation

Tagline: A personal terminal that turns complex care into clear next steps.

Elevator pitch

Families with similar diagnoses often live vastly different lives—because finding the right services is confusing, time-consuming, and opaque. Jayda matches individuals to verified healthcare and social service providers, explains why each match fits (eligibility, documentation, access), and generates click-ready actions (scripts, calls, forms) so people get help faster.

Inspiration

On a soccer sideline, two mothers with similar-needs kids compared notes: one had spent countless hours researching; the other hadn’t had the time or digital skills. With a few shared resources, the second mother’s confidence changed instantly. Jayda exists to replicate that moment at scale—with dignity, security, and speed.

What it does

Smart matching: Structured intake + natural language to map diagnoses, insurance, location, language, documentation, transit constraints, and caregiver availability to vetted providers.

Explainable results: “Why this match” shows eligibility rules, distance, waitlist signals, IDs/forms required, and last-verified source.

Actionable plans: Auto-generated checklists, tap-to-call, secure messages, calendar holds, and printable/PNG summaries for low-connectivity scenarios.

Privacy by design: Minimal collection, client-side redaction, role-scoped views. (See Responsible AI.)

Offline-tolerant UX: Works even with flaky Wi-Fi; summaries can be printed/handed off.

Who benefits: Individuals & caregivers, school counselors, community health workers, urban regional planners, payers/networks seeking equitable access.

Product overview (for judges)

The market spans Medicaid MCOs, health systems, school districts, and community agencies. Our differentiation is an explainable matcher that pairs a rules engine with BAML-defined AI functions so results are transparent, auditable, and quick to act on. Initial GTM: B2B SaaS for agencies/payers + a community edition for families.

Jayda Open Source (OSS): Geospatial Intelligent City Planner

We extended health access + mobility by shipping Jayda OSS — Geospatial Intelligent City Planner, an interactive map + personal terminal that helps cities visualize care access and plan mobility-aware interventions.

How to use (judge-friendly):

Navigation: Pan/zoom/rotate; globe auto-rotates when idle; top-right controls for precision.

Personal Terminal [PeT]: Configure HIPAA/PHI handling and encryption defaults.

Mapbox token: Replace the hardcoded token; for prod, store keys securely (e.g., Supabase).

Interact: Hover buildings to highlight; smooth atmospheric effects.

Quick Action: Generate a 20-minute drive-time isochrone from current location; our MCP tool returns demographics, healthcare facilities, transit networks, POIs, and accessibility metrics within the zone.

Tips: Double-click to zoom; Shift+drag box-zoom; Ctrl+drag rotate; Esc resets globe.

How it works (technical overview)

Stack: Lovable (rapid UI), Supabase (Postgres/Auth/Storage), BAML (deterministic AI functions for entity extraction, eligibility reasoning, routing), plus embeddings for search/reranking.

Data model: Canonical provider schema (diagnoses served, eligibility, insurance, languages, addresses, documentation, referral pathways) with source URL + last-verified timestamp for traceability.

Matching: Rules-first engine (deterministic) augmented by LLM slot-filling for free-text referrals/notes. Every recommendation carries citations + “why this” rationale so caseworkers can trust results.

AI Engines used: GPT-4o, Sonnet 3.5

Responsible AI Remarks:

Models/services: LLMs orchestrated via BAML with typed function specs; no PHI fine-tuning.

Data sources & licenses: Public agency/provider websites and user-submitted entries; each record tracks provenance and verification timestamp.

Safety & ethics: HIPAA-aware flows, least-privilege roles, client-side redaction, on-demand deletion, and auditable access logs; explanations reduce bias and support user control.

HIPAA/PHI note for judges: PHI is individually identifiable health information governed by HIPAA; our app defaults to minimal data capture and de-identification where possible. (References: HHS overview and NIST SP 800-66r2 definition of PHI.) HHS.gov

Challenges we ran into

Function ↔︎ UI mapping (Lovable): Balancing speed of AI-generated layouts with strict interaction patterns (eligibility toggles, progressive disclosure, “why” tooltips).

Data hygiene: Deduping providers, normalizing eligibility language, verifying websites.

Trust & safety: Consent, role-scoped views, and on-device redaction without slowing care.

Timeboxing: Shipping a credible MVP under hackathon constraints while preserving quality where it matters (matching accuracy, privacy, auditability).

Accomplishments we’re proud of

End-to-end loop: Intake → normalized needs → ranked matches → one-tap actions → reminders.

Design & usability: WCAG-aware flows, plain-language copy, quick actions that reduce clicks—strong fit for the Design & Usability criterion.

Technical prowess: Typed schemas, audit logs, explainable outputs mapped to runnable code—aligned with Technical Prowess.

Mobility lens: The OSS 20-min isochrone exposes real access gaps and prioritizes interventions.

What we learned

A 4×100 relay model (clean handoffs between owners) kept us fast and accountable. In-person time accelerates discovery; our long-term goal is a remote team that matches this cadence via explicit rituals, clear ownership, and checklist-driven reviews.

What’s next

Open-source core: Publish matching schemas, BAML functions, and sample provider data for local adaptation.

Verification network: Lightweight contributor workflows to timestamp/verify records.

Caseworker mode: Shared plans, templated scripts, outcome tracking.

IRB-ready research mode: Consent + de-identified analytics to measure time-to-care and enrollment gains.

Ecosystem adapters: Secure connectors for EHRs, payer directories, and community resource platforms.

Testing instructions (judges)

Use the demo credentials in the project page to log in.

Complete the 6–8 minute intake with provided test values.

Review 3–5 ranked matches; click “Why this match” to see eligibility, distance, and documentation.

Click Quick Actions to generate call scripts or a printable action plan.

Open OSS City Planner, hit Quick Action → 20-min isochrone, and scan the returned access metrics.

Built With (tech/APIs)

Lovable, Supabase (Postgres/Auth/Storage), BAML, TypeScript/React, Mapbox GL, Recharts, Tailwind, Vite.

Compliance note (HIPAA)

Certain data types (e.g., PHI, precise location traces, clinical notes) are restricted or de-identified by default. We enforce least-privilege access, client-side redaction, role-scoped views, and auditable logs; privacy is prioritized even when it limits scope. Reference materials for judges on PHI definitions: HHS and NIST SP 800-66r2. HHS.gov

. : Thank you