Aditya-Sivanandan/aceso-medgemma-1.5-4b-it-UD-Q3_K_XL

Project Aceso: MedGemma-1.5-4B-it (UD-Q3_K_XL)

This repository hosts the Tier 1 Intelligence Layer for Project Aceso, optimized for the Med-Gemma Impact Challenge.

Model Description

This is an Unsloth Dynamic (UD) 3-bit quantization (Q3_K_XL) of Google's MedGemma-1.5-4B-it. It is specifically selected for its balance between a tiny memory footprint and high clinical reasoning accuracy on 8GB RAM mobile devices.

🛠️ Credits & Attribution

• Base Model: Google MedGemma-1.5-4B-it.
• Quantization: This model was quantized using the Unsloth Dynamic 2.0 methodology. Special thanks to the Unsloth team for providing the high-efficiency kernels and dynamic layer selection that allow medical models of this scale to run on mobile CPUs.

🩺 Clinical Implementation

This model acts as the "Sovereign Brain" for the Aceso mobile client:

• Task: Generates SBAR (Situation, Background, Assessment, Recommendation) clinical reports from transcripts.
• Safety: Utilizes a Human-in-the-Loop review process where all AI outputs are authorized by a licensed clinician before finalization.

⚖️ Licensing

This model is governed by the Health AI Developer Foundations terms of use. Users must strictly adhere to all ethical guidelines for medical AI as established by Google and the Med-Gemma Impact Challenge.