Speech AI + Agentic Systems Engineer · Türkiye
I build Turkish speech AI that runs locally on a 4 GB GPU, and the self-auditing multi-agent system that develops and maintains it. Türkçe ses yapay zekâsı ve çok ajanlı sistemler; 4 GB kartta, yerel, açık kaynak.
- Turkish voice design · VoxCPM2 on a GTX 1050 Ti, 2.8 GB peak VRAM, Apache-2.0, 65★ / 10 forks
- Word-timed STT (release in progress) · 1 h 12 min of audio in ~7 min on the same GPU
- Speech-to-video pipeline · a 45-minute, 342-sentence documentary produced end to end
- Agent security layer · caught 12/12 test prompt-injection attacks, 0 false alarms on 15 clean samples
Turkish TTS voice design · low-VRAM inference · word alignment & STT · end-to-end speech/video pipelines · multi-agent orchestration with automated checks
Collaboration, consulting, and full-time roles in speech AI. Reach me via GitHub Discussions or X DM.
