---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
tags:
- quantumindssi
- sovereign-ai
- edge-computing
- healthcare-medical-ai
- drug-interactions
- pharmacovigilance
- ddi
- patient-safety
- 09_drug_interaction_predictor
- finetuned
- lora
inference: true
---

# Drug Interaction Predictor

A fine-tuned Small Language Model (SLM) that predicts potential drug-drug interactions from medication lists, including severity, mechanism, evidence, and clinical recommendations.

## Model Details

| Attribute | Value |
|-----------|-------|
| **Developer** | QuantumIndSSI Ltd |
| **Base Model** | ./base_model |
| **Architecture** | Transformer decoder (causal LM) |
| **Fine-tuning Method** | LoRA (Low-Rank Adaptation) |
| **LoRA Rank** | 16 |
| **LoRA Alpha** | 32 |
| **License** | apache-2.0 |

## Intended Use

- Automated drug-drug interaction screening from medication lists
- Clinical decision support for pharmacists and prescribers
- Patient safety and pharmacovigilance workflows
- Edge deployment on Victron and other constrained hardware

## Training Data

- 100,000+ synthetic drug-drug interaction examples
- Drug classes: anticoagulants, statins, antibiotics, antidepressants, PPIs, NSAIDs, opioids, immunosuppressants, and more
- Interaction types: pharmacokinetic (CYP inhibition/induction, P-gp), pharmacodynamic (additive toxicity, QT prolongation, bleeding)
- Severity levels: Critical, High, Medium, Low

## Usage

```python
model_id = "QuantumindSSI/09_drug_interaction_predictor"
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)

prompt = """Medications: Warfarin, Aspirin, Omeprazole. Analyze for interactions."""
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=512)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

## Limitations

- Not a substitute for clinical pharmacologist review
- Synthetic training data may not capture all real-world interactions
- English only
- Always verify with FDA labels, clinical guidelines, and drug databases

## Hardware Requirements

| Target | RAM | Notes |
|--------|-----|-------|
| Cloud GPU | 4GB | FP16 inference |
| Workstation | 3GB | INT8 quantized |
| Victron Edge | 2-3GB | INT8/INT4 quantized, CPU |

## Citation

```bibtex
@misc{09_drug_interaction_predictor,
  title={Drug Interaction Predictor},
  author={QuantumIndSSI Ltd},
  year={2026},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/QuantumindSSI/09_drug_interaction_predictor}}
}
```

## Contact

- GitHub: https://github.com/QuantumindSSI
- HuggingFace: https://huggingface.co/QuantumindSSI