I am deeply committed to the intersection of Deep Learning, Ethnomusicology, and Digital Preservation. My lifelong mission is to build computational frameworks that protect and celebrate the rich, microtonal traditions of Somali music through modern AI architectures.
I am currently leading research at the Qaraami-Gen-AI Lab, where we develop tools for:
- Automatic Archival Restoration: Using deep source separation to recover clean Oud stems from vintage Somali cassettes.
- Microtonal Music Transformers: Designing custom attention mechanisms for the 5-note Somali pentatonic scale to prevent "Western-bias" in symbolic generation.
- Interactive Jam Spaces: Prototyping Human-AI improvisation systems inspired by interaction-driven generative design.
- Languages: Python (Research), TypeScript/JavaScript (Web UI/UX).
- Deep Learning: PyTorch, TensorFlow, Transformers, Social Reinforcement Learning.
- Audio Engineering: Librosa, Demucs, Spleeter, FFmpeg, Constant-Q Transform (CQT).
- Data & Backend: MusicXML, MIDI, FastAPI, PostgreSQL (Neon), Next.js.
- 🎙 Pre-processing: Successfully isolated high-fidelity Oud tracks from historical mono-recordings using neural source separation.
- 🎼 Transcription: Developing a personalized pitch-tracking system that recognizes neutral notes often "crushed" by standard MIDI systems.
- 🌍 Community: Building an open-source corpus of machine-readable Somali microtonal music.
- Research Portal: qaraamigen-ai.vercel.app
- Organization: Qaraami-Al-Research
- LinkedIn: khalidibrahimabdi1
- Twitter: @khalid_ibrahim
"Training the system to hear the silence between the notes."