AI that thinks and works like your expert. eqlx.ai · İstanbul
eqlx builds AI for work that takes expertise, for financial institutions and fintechs, and for industrial and service companies:
- Machine-learning models score: probabilities, rankings, forecasts, and the action each band triggers inside the institution's thresholds.
- Small language models trained for one domain and task reason: they read like your expert, call the right tool, and write in your format.
- Agentic AI systems act: they prepare each case and act only inside the limits the institution writes.
Your expert decides. Each part works on its own; together where the work needs it. Everything runs in the customer's environment and is handed over running.
Open-source tools: eqlx.org
Deterministic tools for AI agents in credit and finance. The agent decides what to check and when; the tool computes the number, the same way every time, checked against public sources.
| Tool | What it computes |
|---|---|
| eqlx-credit | PD model validation and Basel IRB risk weights |
| eqlx-ebelge | e-Fatura and e-Defter files read into clean data and a trial balance, with checks |
| eqlx-aml | Money-laundering patterns in transfer data, with the rows behind each alert |
| eqlx-ifrs9 | IFRS 9 expected credit loss, with the rule behind every stage |
| eqlx-bankmath | Bank product arithmetic in Türkiye, exact to the kuruş |
Each is Apache-2.0, on PyPI, and ships a read-only MCP server where an agent needs one.
- Hugging Face: eqlx-ai: Turkish routing and speech models
- korkut-gpt: a small Turkish language model built from scratch, with a lesson series on YouTube
Field notes from the work, in English and Turkish:
- Agentic RAG or plain RAG?
- Workflow or agent? Ask whether the next step changes
- Scorecard or gradient boosting? Let held-out data decide
- A LoRA adapter or a full fine-tune?
Use cases · Glossary · Decision Blueprint
Equalize the gap.