Health Report & Lifestyle Advisor

Hacko-ADK: AI-powered health platform with Google Gemini, seamless intake, lab-report analysis, real-time chat, and personalized summaries—secure, scalable, ideal for providers and wellness startups.

Comment

Inspiration

Imagine a digital health assistant that not only understands your medical reports but also helps you set goals, find the right doctors, and delivers personalized lifestyle and dietary guidance —all in one seamless conversation. That vision drove the project: a multi-agent AI system designed to make healthcare more accessible, actionable, and human-centred.

Key goals:

  • Make medical information understandable for everyone.
  • Provide actionable, personalized recommendations.
  • Help users take real steps toward better health.
  • Seamlessly connect users to real-world resources (like doctors and events).

What it does

The system empowers users to:

  1. Upload a lab report and receive a plain-language summary of findings.
  2. Request tailored diet or lifestyle plans, delivered as downloadable PDF guides.
  3. Locate nearby doctors when critical results appear.
  4. Keep all conversation-context in one chat

How we built it

Component Role
Manager Agent Orchestrates the conversation and maintains session state.
Intake Agent Collects basic demographics and health history.
Explanatory Agent Parses lab PDFs, flags normals vs. criticals, suggests next steps.
Dietary Agent & Lifestyle Agent Craft personalized meal plans and habit road-maps, output via PDF Creator Tool.
Goal Setter Agent Schedules reminders and recurring activities via calendar APIs.
Tools PDF Creator, Calendar utilities, Doctor Finder for nearby specialists.

Workflow example:

  1. User uploads a report → Manager → Explanatory Agent.
  2. User asks for diet advice → Manager → Dietary Agent → PDF Creator Tool.
  3. Critical marker detected → Explanatory Agent → Doctor Finder Tool.

Challenges we ran into

  • Messy PDFs: Labs use dozens of report templates, forcing hybrid DocAI + regex extraction.
  • Context hand-off: Keeping every agent up-to-date without bloating prompts demanded a shared, lightweight session state.
  • Privacy & compliance: Handling PHI securely (encrypted temp storage, signed URLs, auto-purge) added early architectural constraints.
  • Multilingual clarity: Ensuring explanations stayed clear across English, Hindi, and Tamil required careful prompt-engineering and language-detection.

Accomplishments that we're proud of

  • Built a fully modular multi-agent demo that anyone can extend without touching core logic.
  • Dropped the “What does this value mean?” burden by turning raw results into instant plain-English summaries and action plans.
  • Open-sourced the system as one of the first public Google-ADK healthcare examples, helping others bootstrap similar solutions.

What we learned

  • Small, role-specific agents > monoliths. Clear boundaries speed iteration and debugging.
  • Session state is everything. A shared key-value store beats passing entire histories in every prompt.
  • Explain, then prescribe. Users trust personalised reasoning more than generic risk scores.
  • Bake compliance in early. Adding security layers later is painful.

What's next for Health Report & Lifestyle Advisor

  1. Wearable & CGM data fusion for real-time coaching.
  2. FHIR export to integrate directly with hospital EMRs.
  3. Trend-watch agent that flags creeping risks (e.g., rising HbA1c).
  4. Community challenges to boost adherence through social accountability.
  5. Regulatory readiness — pursuing ISO 13485 & CDSCO clearance for an India launch.

Built With

Share this project:

Updates