SimplyFI — AI Agents for Banking
Intelligent automation for Customer Engagement, Compliance, Lending, and Trade Finance.
Models, datasets, and Spaces for document understanding, risk, and decisioning — production-first and audit-ready.
---
## What we publish
- **Models**: OCR/IE, invoice understanding, entity linking (KYC/AML), credit signals, dialogue agents (RAG+RLHF).
- **Datasets**: Synthetic & anonymized corpora for invoices, SWIFT/LC patterns, trade compliance, FAQs.
- **Spaces**: Demo apps for **document triage**, **explainable decisions**, and **agentic workflows**.
- **Notebooks**: End-to-end examples — from fine-tuning to human feedback (RLHF) and safe deployment.
> Prefer **bank-grade** controls? We support **gated access**, **EULAs**, and **private org Spaces** for sensitive assets.
---
## Highlights
- **SIMBA** — our production platform for **Trade Finance & Banking AI**: OCR ➜ extraction ➜ validation ➜ decisioning
- **Agentic Workflows** — multi-tool agents with grounded retrieval, guardrails, and human-in-the-loop review
- **Compliance by design** — redaction, PII handling, lineage, prompt/trace logging, and model cards with usage guidance
---
## Explore
### Featured models
| Purpose | Model | Tasks | Notes |
|---|---|---|---|
| Invoice Understanding | [`SimplyFI/invoice-extractor`](https://huggingface.co/SimplyFI/invoice-extractor) | token-classification, table-qa | line-item, taxes, totals, vendor |
| Trade Compliance NER | [`SimplyFI/trade-ner-compliance`](https://huggingface.co/SimplyFI/trade-ner-compliance) | token-classification | sanctions terms, ports, HS codes |
| Banking Assistant (SFT) | [`SimplyFI/banking-assistant-sft`](https://huggingface.co/SimplyFI/banking-assistant-sft) | text-generation | safe, concise answers with citations |
| Risk Signals | [`SimplyFI/risk-signal-classifier`](https://huggingface.co/SimplyFI/risk-signal-classifier) | sequence-classification | escalations & review routing |
> Replace links above with your actual repos (or keep as placeholders until published).
### Example: quick inference
```python
from transformers import AutoModelForTokenClassification, AutoTokenizer, pipeline
repo = "SimplyFI/invoice-extractor" # update if different
tok = AutoTokenizer.from_pretrained(repo)
mdl = AutoModelForTokenClassification.from_pretrained(repo)
ner = pipeline("token-classification", model=mdl, tokenizer=tok, aggregation_strategy="simple")
text = "Invoice 9081 from Alpha Plastics Pvt Ltd, Total ₹1,24,560 due on 2025-11-15"
print(ner(text))