Spaces:
Configuration error
Configuration error
|
Download docs/MENTOR_DEMO.md from Sivaneshakumar/sanjeevani-api: direct link, hf CLI and curl.
- Browser
- Download file 4.31 kB
-
https://huggingface.co/spaces/Sivaneshakumar/sanjeevani-api/resolve/main/docs/MENTOR_DEMO.md
- Command line
-
hf download hf://spaces/Sivaneshakumar/sanjeevani-api/docs/MENTOR_DEMO.md
-
curl -L -o MENTOR_DEMO.md https://huggingface.co/spaces/Sivaneshakumar/sanjeevani-api/resolve/main/docs/MENTOR_DEMO.md
4.31 kB
| # SanjeevaniAI β Mentor Presentation & Technical Demonstration Guide | |
| **Duration**: 10β15 Minutes | |
| **Target Audience**: Technical Mentors, Senior Architects, Faculty Evaluators | |
| --- | |
| ## π― Executive Pitch (1 Minute) | |
| > *"SanjeevaniAI is an industry-grade healthcare intelligence and clinical decision-support platform designed to assist healthcare professionals and patients with biomedical named entity recognition, medical report analysis, and structured conversational AI consultation. It runs a fine-tuned RoBERTa-large BC5CDR model completely locally on disk without third-party NER dependencies, enforces strict non-diagnostic clinical safety boundaries, and provides an immutable audit trail."* | |
| --- | |
| ## π Demonstration Flow | |
| ### Stage 1: Local Neural Model Verification (3 Minutes) | |
| 1. Navigate to `/ner` (Biomedical NER Visualizer). | |
| 2. **Show the Mentor**: | |
| - Status badge confirms model loaded locally from `D:\SanjeevaniAI\models\bc5cdr-ner`. | |
| - Hardware device indicates active acceleration (`cuda:0` / `CPU`). | |
| 3. Click the preset: **"Diabetes & Hypertension (Standard Case)"**. | |
| 4. Click **"Run BC5CDR Inference"**. | |
| 5. **Key Talking Points**: | |
| - Notice the exact token highlight: `metformin` and `lisinopril` tagged in **emerald** as `CHEMICAL`, `type 2 diabetes mellitus` and `hypertension` tagged in **rose** as `DISEASE`. | |
| - Sub-word token BPE reconstruction correctly handles multi-word entities and punctuation. | |
| - Inference completes in under **20 milliseconds** with confidence scores $>99.8\%$. | |
| --- | |
| ### Stage 2: Medical Report Ingestion & Structured Extraction (3 Minutes) | |
| 1. Navigate to `/reports`. | |
| 2. Inspect the pre-loaded synthetic report: `Metabolic_Panel_Report_2026.txt`. | |
| 3. Click **"View Analysis"** (`/reports/[id]`). | |
| 4. **Key Talking Points**: | |
| - Shows SHA-256 fingerprinting for tamper-evidence. | |
| - Shows automated extraction of key findings, aggregated conditions, and pharmacological agents. | |
| - Non-diagnostic summary language: framed strictly as educational decision support. | |
| --- | |
| ### Stage 3: Conversational Clinical AI & Emergency Triage (4 Minutes) | |
| 1. Navigate to `/assistant`. | |
| 2. **Standard Query Demo**: | |
| - Ask: *"Why is metformin taken with meals and what are typical gastrointestinal side effects?"* | |
| - Show structured response output: | |
| - π Clinical Summary | |
| - π‘ Considerations (extended-release formulations, Vitamin B12 absorption) | |
| - β Questions to ask your doctor | |
| - β οΈ Medical disclaimer | |
| 3. **Emergency Red-Flag Triage Demo**: | |
| - Click the prompt: *"I am having severe crushing chest pain, shortness of breath, and left arm numbness."* | |
| - **Show the Mentor**: | |
| - System triggers the **emergency triage heuristic engine**. | |
| - Prominently displays the **Red Emergency Alert Banner** directing the user to call 911 / 112 / 108 or proceed to the nearest emergency department immediately. | |
| --- | |
| ### Stage 4: Patient Longitudinal Health Profile & Timeline (2 Minutes) | |
| 1. Navigate to `/profile`. | |
| 2. Show the real-time BMI calculator, allergies list, chronic conditions, and current medications. | |
| 3. Explain how this clinical profile is automatically injected as context into the AI consultation assistant. | |
| 4. Navigate to `/history` to show the chronological activity log with entity counts and timestamps. | |
| --- | |
| ### Stage 5: System Telemetry & Security Audit Trail (2 Minutes) | |
| 1. Click **"Demo Roles"** in the top navigation bar and select **"Admin"** (or log in as `demo.admin@sanjeevani.ai` / `DemoAdmin2026!`). | |
| 2. Navigate to `/admin`. | |
| 3. **Show the Mentor**: | |
| - Real-time platform aggregates: Total Users, Processed Documents, Extracted Biomedical Entities. | |
| - Live hardware status of the local RoBERTa model. | |
| - Tamper-evident Security Audit Log recording logins, report uploads, and consultation events. | |
| --- | |
| ## π Pre-Seeded Demonstration Accounts | |
| | Role | Email | Password | Purpose | | |
| | :--- | :--- | :--- | :--- | | |
| | **Patient** | `demo.patient@sanjeevani.ai` | `DemoPatient2026!` | Standard patient portal, health profile, medical documents | | |
| | **Doctor** | `demo.doctor@sanjeevani.ai` | `DemoDoctor2026!` | Clinician decision support and report review | | |
| | **Administrator** | `demo.admin@sanjeevani.ai` | `DemoAdmin2026!` | System telemetry, model health, and security audit logs | | |