Independent ML/NLP researcher based in Faisalabad, Pakistan, specializing in:
- 🤖 Large Language Models — fine-tuning, GRPO, QLoRA, inference on H100 GPUs
- 🔍 Retrieval-Augmented Generation (RAG) — enterprise-grade pipelines
- 👁️ Computer Vision — 3D segmentation, physics-informed neural networks
- 📊 Competitive ML — ensemble methods, calibration, generalization theory
| Competition | Medal | Score |
|---|---|---|
| 🥉 Vesuvius Challenge — Surface Detection | Bronze | Dice Score 0.8607 |
| 🥉 AIMO3 — Mathematical Reasoning with LLMs | Bronze | 42 / 50 · GPT-OSS-120B · H100 GPU |
1. Calibrated Ensemble Regularization for Competitive ML (Under Review — TMLR, May 2026)
"Beyond the Public Leaderboard: When Does Calibrated Ensemble Regularization Reduce the Generalization Gap in Competitive Machine Learning?" Proposes the GACE method for reducing the train–test generalization gap in competitive ML settings. → OpenReview
2. Enterprise RAG & Scalable ML Deployment (Zenodo, 2025)
End-to-end framework for production RAG pipelines and scalable ML deployment in enterprise environments. → DOI: 10.5281/zenodo.17195112
3. SparsePhysNet — Physics-Informed Sparse Attention Transformer (Zenodo, 2025)
Novel transformer integrating physics-informed constraints with sparse attention for scientific computing tasks. → DOI: 10.5281/zenodo.19365914
| Platform | Link |
|---|---|
| 🤗 HuggingFace | huggingface.co/sibi-ai |
| 📊 Kaggle | kaggle.com/hasib007 |
| linkedin.com/in/hasib007 | |
| 🐙 GitHub | github.com/sibi-ai |
| haseb_ahmad@yahoo.com |