Skip to main content
Back to challenges
Challenge

Build Multilingual Voice Health Assistant with AutoGen and Cartesia

Rural citizen healthcare access requires intuitive natural language interfaces across regional Indian languages. Build a multi-agent conversational assistant using AutoGen paired with Cartesia text-to-speech models to process Ayush traditional medicine queries with sub-second speech synthesis.

Business OperationsHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Develop a multi-agent conversational healthcare counselor using AutoGen multi-agent orchestration and Cartesia high-speed voice streaming.

Delivery guide

How work is evaluated

Evaluation

Validates multi-agent conversation correctness and Cartesia audio stream synthesis quality.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

safety_compliance_check

Verifies output contains mandatory disclaimer for acute symptoms

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

latency_ms

Total response time including TTS synthesis • target: 800 • range: 0-2000

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Orchestrate collaborative specialized agents using Microsoft AutoGen for health triage

  • Integrate Cartesia ultra-low latency TTS API for real-time voice response generation in Hindi and Tamil

  • Verify medical entity extraction accuracy against WHO Ayush traditional medicine standards

Resources and assets

Reference links and supporting material

Dataset notes

Parallel dataset of citizen health queries in 4 Indian languages mapped to standardized Ayush remedies.

Challenge setup

Validates multi-agent conversation correctness and Cartesia audio stream synthesis quality.

Challenge input

JSON string containing citizen query in Hindi/Tamil text

AutoGen

Coordinates multi-agent safety checks and domain reasoning

Cartesia

Delivers real-time low latency text-to-speech

Evaluated output

JSON containing recognized medical concept, remedies, and audio stream URL

Evaluation checks
  • Verifies output contains mandatory disclaimer for acute symptoms
  • Total response time including TTS synthesis • target: 800 • range: 0-2000
Configured evaluation evidence
  • Benchmark: Voice-Health-Bench-India
  • Latency Ms target: 800
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • Cartesia · Required
  • AutoGen · Optional
Observation
  • Galileo · Optional
Reward / Eval
  • Galileo · Optional

Evaluation contract

  • safety_compliance_check · Weight 1
  • latency_ms · Weight 1

Metrics

  • latency_ms · Target 800

Benchmarks

  • Voice-Health-Bench-India
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON string containing citizen query in Hindi/Tamil text

    Output: JSON containing recognized medical concept, remedies, and audio stream URL

    Execution harness: Python

Evaluation tools

  • AutoGen · Agent Orchestration
  • Cartesia · Voice Synthesis

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build Multilingual Voice Health Assistant with AutoGen and Cartesia