🏗️ Building on HF

Göktuğ Düşünen PRO

GoktugD

11 3

https://werea.co

AI & ML interests

Head of AI @ Werea · AI/ML Engineer AI/ML Engineer building end-to-end intelligent systems across LLMs, NLP, Computer Vision, RAG, Retrieval, Agents, and Applied AI. From data and model development to fine-tuning, evaluation, optimization, deployment, and production AI systems.

Recent Activity

posted an update about 10 hours ago

🇹🇷 Our 1B Turkish OCR model achieved 0.15% character error rate. But there's a catch. That result came from a controlled evaluation. Real-world documents are a different challenge. Blurry phone photos. Damaged receipts. Complex tables. Handwritten annotations. Low-resolution scans. These are the cases where OCR systems can struggle. We built Werea-DocOCR-1B, an open-source model specialized for Turkish enterprise documents. Our published evaluation: 📄 12 document categories 🧪 72 held-out synthetic test pages 📊 0.15% average CER 🖥️ Fine-tuned on a single RTX 3090 🔓 Apache 2.0 But here's what interests me more than another benchmark: Can the Hugging Face community find documents that break our model? 🔥 THE TURKISH OCR CHALLENGE I'm looking for difficult Turkish documents that expose real weaknesses. If you work with OCR, document AI, or multilingual models, I'd love your input. We'll use the most interesting failure cases to improve our evaluations and share what we learn. No inflated claims. No hiding failures. Just open models and measurable progress. 🤗 Model: https://huggingface.co/Werea-co/Werea-DocOCR-1B 📚 Dataset: https://huggingface.co/datasets/Werea-co/werea-tr-doc-ocr-enterprise-v2 🇹🇷 Built in Türkiye. Open to everyone. Follow @GoktugD and @Werea-co to see the results of the challenge and our next experiments. 💬 Question: What's the hardest document type you've ever tried to extract text from? #OCR #DocumentAI #OpenSourceAI #HuggingFace #TurkishAI

posted an update 2 days ago

🚀 **Do We Really Need Billion-Parameter AI Models for Every Task?** The AI industry is obsessed with scale. Bigger models. More parameters. More GPUs. More compute. But what if we're overlooking something important? **Efficiency matters just as much as scale.** I've been working on open-source AI models focused on Turkish NLP, document understanding, OCR, and specialized language tasks. My goal is simple: Build AI models that are **smaller, faster, more accessible, and genuinely useful.** 🔹 Small language models optimized for specific tasks 🔹 Efficient inference without expensive GPU infrastructure 🔹 Open-source models and datasets 🔹 Better AI support for underrepresented languages 🔹 Practical applications instead of benchmark hype I believe the future of AI won't be dominated exclusively by massive foundation models. It will also belong to **small, specialized models that solve real problems efficiently.** And that's exactly what I'm building. I'm sharing my experiments, models, datasets, and research openly on Hugging Face. 🤗 **Explore my work:** https://huggingface.co/GoktugD 💬 **Here's a question for the AI community:** Would you rather deploy a specialized 300M-parameter model that performs your task well, or a general-purpose 8B model that requires significantly more resources? I'm genuinely interested in hearing your experiences. **Let's make powerful AI accessible to everyone. 🌍** #OpenSourceAI #SmallLanguageModels #MachineLearning #HuggingFace #EfficientAI

new activity about 1 month ago

Werea-co/Werea-TR-TextRestore:Demo for this model on Spaces

View all activity

Organizations

Posts 9

Post

🇹🇷 Our 1B Turkish OCR model achieved 0.15% character error rate. But there's a catch.

That result came from a controlled evaluation.

Real-world documents are a different challenge.

Blurry phone photos.
Damaged receipts.
Complex tables.
Handwritten annotations.
Low-resolution scans.

These are the cases where OCR systems can struggle.

We built Werea-DocOCR-1B, an open-source model specialized for Turkish enterprise documents.

Our published evaluation:

📄 12 document categories
🧪 72 held-out synthetic test pages
📊 0.15% average CER
🖥️ Fine-tuned on a single RTX 3090
🔓 Apache 2.0

But here's what interests me more than another benchmark:

Can the Hugging Face community find documents that break our model?

🔥 THE TURKISH OCR CHALLENGE

I'm looking for difficult Turkish documents that expose real weaknesses.

If you work with OCR, document AI, or multilingual models, I'd love your input.

We'll use the most interesting failure cases to improve our evaluations and share what we learn.

No inflated claims.
No hiding failures.
Just open models and measurable progress.

🤗 Model:
Werea-co/Werea-DocOCR-1B

📚 Dataset:
Werea-co/werea-tr-doc-ocr-enterprise-v2

🇹🇷 Built in Türkiye. Open to everyone.

Follow @GoktugD and @Werea-co to see the results of the challenge and our next experiments.

💬 Question: What's the hardest document type you've ever tried to extract text from?

#OCR #DocumentAI #OpenSourceAI #HuggingFace #TurkishAI

Post

🚀 **Do We Really Need Billion-Parameter AI Models for Every Task?**

The AI industry is obsessed with scale.

Bigger models. More parameters. More GPUs. More compute.

But what if we're overlooking something important?

**Efficiency matters just as much as scale.**

I've been working on open-source AI models focused on Turkish NLP, document understanding, OCR, and specialized language tasks.

My goal is simple:

Build AI models that are **smaller, faster, more accessible, and genuinely useful.**

🔹 Small language models optimized for specific tasks
🔹 Efficient inference without expensive GPU infrastructure
🔹 Open-source models and datasets
🔹 Better AI support for underrepresented languages
🔹 Practical applications instead of benchmark hype

I believe the future of AI won't be dominated exclusively by massive foundation models.

It will also belong to **small, specialized models that solve real problems efficiently.**

And that's exactly what I'm building.

I'm sharing my experiments, models, datasets, and research openly on Hugging Face.

🤗 **Explore my work:**
@GoktugD

💬 **Here's a question for the AI community:**

Would you rather deploy a specialized 300M-parameter model that performs your task well, or a general-purpose 8B model that requires significantly more resources?

I'm genuinely interested in hearing your experiences.

**Let's make powerful AI accessible to everyone. 🌍**

#OpenSourceAI #SmallLanguageModels #MachineLearning #HuggingFace #EfficientAI