Computer Engineering graduate from Marmara University, specializing in Artificial Intelligence, NLP, Speech Recognition (ASR), and Large Language Models (LLMs). I build production-ready, real-time AI systems with Python, FastAPI, Kafka, Elasticsearch, MongoDB, and Docker.
- 🔬 Led TÜBİTAK 2209-A–funded Turkish Medical ASR research (graduation project)
- 🏥 Work spans medical AI, multimodal deepfake detection, and privacy-preserving LLM architectures
- 🤗 Published Turkish NLP & speech datasets and models on Hugging Face
- 🚀 Co-founder of PulsarMetric, a health-tech startup (ECG analysis & training)
- 📫 Reach me at y.tuncer1004@gmail.com · Open to job opportunities
| Area | What I build |
|---|---|
| 🎙️ Speech Recognition (ASR) | Turkish & medical speech-to-text, diarization, audio-analysis pipelines (WhisperX, NeMo) |
| 📝 Turkish NLP | NER, text normalization, fine-tuned BERT, ASR post-processing |
| 🏥 Medical AI | Privacy-preserving medical LLMs, cardiovascular risk, health-tracking systems |
| 🤖 LLMs & Computer Vision | Distributed/secure LLM inference, multimodal deepfake detection |
| Project | Description |
|---|---|
| turkmedstt | Turkish medical ASR (TÜBİTAK 2209-A). Two-stage LoRA fine-tuned Whisper Large V3 — −34.7% WER, −58.6% CER vs. base; 20-model benchmark + HF leaderboard/demo |
| multimodal-deepfake-detection | Multi-task deepfake detection on FakeAVCeleb — Xception + Wav2Vec 2.0 + lip-sync via late fusion (any-AUC 0.9994) |
| privacy-preserving-medical-llm | Secure medical LLM for cardiovascular risk — Llama 3.1 8B, split learning, AES-256-GCM / RSA-4096, JWT, audit logging |
| Speech-to-Text-Api | Enterprise ASR API — WhisperX / NeMo / Parakeet, diarization, emotion, Kafka + MongoDB + Elasticsearch |
| Ner-Ai-Project | Production Turkish NER + synthetic data, fine-tuned BERT, FastAPI UI |
| MediGuard-ESP32 | IoT smart medicine reminder & adherence tracker (ESP32, Wokwi) |
Languages
AI / ML
Backend & Data
Infrastructure & Security
Let's build something meaningful. 🚀