Efficient intelligence. Careful experiments. Open research.
DeltaTune Lab is an independent AI research lab founded by Pakistan-based researchers Rao Muhammad Dayan Atif and Syed Muhammad Sarim Ahsan. The lab explores practical ways to adapt, improve, and run modern AI systems with less unnecessary overhead.
- Efficient fine-tuning and model adaptation
- Reinforcement learning and post-training
- Compact model reasoning
- Model optimization and distillation
- GPU systems, kernel fusion, and memory efficiency
Our research is guided by three questions:
- How much really needs to change?
- What makes a small model reason better?
- Where is the compute going?
Co-founder focused on AI systems, model optimization, and research implementation.
Co-founder focused on efficient AI research, model adaptation, and practical machine learning systems.
Both founders are connected with Sir Syed University.
DeltaTune Lab is building its research foundation in the open. Published papers, models, datasets, and implementation details will be added as they become ready for public release.
Research diagrams and early ideas in this repository are conceptual unless explicitly marked as evaluated results.
DeltaTune Lab is at the beginning of its public research journey. We prefer clear questions, honest limitations, reproducible work, and useful releases over unsupported claims.
