Skip to content
View Abu-Sameer-66's full-sized avatar
23 followers · 20 following

Block or report Abu-Sameer-66

Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Abu-Sameer-66/README.md
Typing SVG

     

👨‍💻 Engineering Philosophy

I am an AI/ML Engineer, Data Scientist, and Open Source Contributor. My foundation is built on Python, SQL, and System Architecture.

Currently, I am focused on LLM Optimization—making massive models run on consumer hardware. I don't just use libraries; I debug the mathematics behind them.

  • 🔭 Current Focus: Native Integration of Instruction-Tuned LLMs into DeepChem.
  • 💡 Core Skill: Fixing CUDA OOM & NaN Loss in resource-constrained environments.
  • ⚡ Goal: Delivering high-impact code for GSoC 2026.

🛠️ Tech Stack

Languages

DL Stack



Data Stack

Tools

🏆 Featured Research & Projects

Project Domain Technical Impact Stack
Mistral7B-BACE-Inhibitors 🧬 Molecular AI Scaffold-Split Generalization. Captured peak ROC-AUC of 0.8034 using LoRA ($r=8$) to prevent SMILES memorization and regularize the adapter. DeepChem Mistral-7B PEFT
Mistral7B-Tox21-Optimization 🧪 Cheminformatics Numerical Stability Engineering. Resolved NaN loss crashes via fp32 upcasting; reached 0.72 ROC-AUC with 8x dynamic oversampling. HuggingFace BitsAndBytes LoRA
ChemLLM-Tox-OLMo 🧠 Multi-Task AI Instruction Tuning. Adapted OLMo-7B for simultaneous multi-task prediction across 12 toxicological benchmarks using instruction-tuning datasets. OLMo-7B DeepChem PyTorch
CogniPath Analytics 🏫 EdTech AI Predictive Intelligence. OOP-based analytics engine forecasting student burnout using behavioral telemetry and XGBoost pipelines. XGBoost OOP Scikit-Learn
PropVal-AI Engine 🏠 PropTech Enterprise Valuation Model. High-variance AVM (Automated Valuation Model) designed for real estate market volatility with robust feature engineering. Gradient Boosting Pandas

📈 Engineering Metrics




Pinned Loading

  1. SpamGuard-AI-Threat-Detection SpamGuard-AI-Threat-Detection Public

    Enterprise-grade SMS Phishing & Spam Detection System utilizing NLP and Probabilistic Machine Learning.

    Jupyter Notebook

  2. PropVal-Property-Valuation--AI-Real-Estate-Engine PropVal-Property-Valuation--AI-Real-Estate-Engine Public

    Production-grade Automated Valuation Model (AVM) utilizing Gradient Boosting to estimate real estate asset values with high precision.

    Jupyter Notebook 1

  3. CogniPath-Analytics-Engine CogniPath-Analytics-Engine Public

    AI-driven EdTech analytics engine optimizing academic outcomes via predictive modeling and behavioral analysis.

    Jupyter Notebook

  4. ChemLLM-Tox-OLMo ChemLLM-Tox-OLMo Public

    Fine-tuning OLMo-7B with QLoRA & DeepChem for Molecular Toxicity Prediction on the Tox21 dataset.

    Python

  5. Mistral7B-Tox21-Molecular-Optimization Mistral7B-Tox21-Molecular-Optimization Public

    Native fine-tuning of Mistral-7B on the Tox21 dataset using LoRA and 4-bit quantization, achieving a competitive 0.72 ROC-AUC.

    Jupyter Notebook

  6. Mistral7B-BACE-Generalization-Study Mistral7B-BACE-Generalization-Study Public

    Optimizing Mistral-7B for BACE-1 inhibitor prediction using Low-Rank Adaptation (r=8) to capture peak generalization on scaffold-split molecular data. Achieved 0.8034 ROC-AUC. Built for DeepChem GS…

    Jupyter Notebook