Athena Coding POC

Mobile Coding Flow

Mobile-first workflow: pick patient, review note, generate coded draft, compare against truth.

Patient

Encounter

Synthetic doctor note

Pick patient + encounter to auto-build the blind payload.

Vertex Draft Settings

Model

Thinking level

Waiting for blind payload.

Compare to Real Codes

Agent output

No comparison yet.

Run compare to see a direct explanation of where your output differs from the algorithm grading.

Auto follow-up to ask why

AI reasons for each code

Nuance dictionary (lower confidence edges)

Ask follow-up

No follow-up asked yet.

1 Select Synthetic Patient

Load a realistic synthetic patient with coherent clinical history and encounters to serve as the demo case.

Supports the built-in demo patient JSON format and Synthea FHIR bundle exports. Imported patients persist locally and appear in the same picker.

Seed

Gender

Min Age

Max Age

State

City

Generation uses a real Synthea runtime when configured locally or available through Docker. Seed, gender, and age range help produce comparable patients.

On-demand Synthea generation has not run yet.

Patient library has not been loaded yet.

Selected Patient

Selected Encounter

1A. Athena Sandbox Authentication

Connect to an Athena sandbox using Client Credentials.

Base URL

Practice ID

Scope

Client ID

Client Secret

No token loaded yet.

No Athena capability probe has run yet.

No Athena write probe has run yet.

View Athena Probe JSON

View Athena Write Probe JSON

2A. Encounter Discovery

Find appointments with clinical activity in the specified date range.

Parsed Department

Active Departments

View Raw Department Responses

Departments JSON

Activity JSON

Dept ID

Provider ID

Start Date

End Date

No appointments loaded.

Select	Enc ID	Appt ID	Date	Patient	Status

3A. Find Best Demo Encounter

No scan run yet.

Select	Enc ID	Score	Svcs	Diag	Closed Note

View Scan JSON

Selected Encounter ID

Candidate Doctor Note

2 Build Blind Payload

This payload is handed to the coding agent. It contains clinical narrative but no billing truth.

Exact note text being sent to the agent

Suggested Agent Prompt

Generated Payload (JSON)

3 Vertex AI Draft Generation

Run the blind payload through Vertex AI. Models are discovered live from the selected project and location so you can compare what Vertex actually exposes there.

Project ID

Region

Model

Thinking Level

Applied only when the selected model supports thinking.

Custom Vertex Authentication

Toggle this on to override default server credentials. Settings are cached in local storage and can be read/saved via endpoint.

Enable custom auth

The app loads the live Vertex model list for the selected region and also merges in global models when Vertex only exposes them there.

Vertex AI configuration has not been loaded yet.

View Exact Prompt Sent to Vertex

View Raw Vertex Response

4 Evaluate Agent Output

Build the held-back truth payload and compare it against the agent's response.

Source of Truth

Agent Output

No comparison has been run yet.

Athena Coding POC

Sign in to access the live Codebill demo

What happens after login

Mobile Coding Flow

Vertex Draft Settings

Compare to Real Codes

Auto follow-up to ask why

AI reasons for each code

Ask follow-up

1 Select Synthetic Patient

Patient Demographics

Active Conditions

Active Medications

1A. Athena Sandbox Authentication

2A. Encounter Discovery

3A. Find Best Demo Encounter

1B. Blue Button Claims Truth

2 Build Blind Payload

3 Vertex AI Draft Generation

Custom Vertex Authentication

4 Evaluate Agent Output

Source of Truth

Agent Output

Comparison Results

Code Difference Snapshot

CPT / Procedures

ICD-10 / Diagnoses