I’m a Data Scientist & Machine Learning Engineer with strong expertise in deploying AI systems at scale — both in cloud and offline setups. Currently serving as Assistant Director of Data Analytics (AI Lead) at NADRA, Pakistan.
I specialize in:
- Generative AI & Large Language Models (LLMs)
- Federated & Distributed Learning
- Real-time OCR, Speech-to-Text & Biometric Intelligence
- Scalable Deployments on AWS, Azure, and On-Prem
🧠 My academic background includes an MS in Systems Engineering from NUST, and I am a certified professional from IBM, Microsoft, NVIDIA, and AWS.
- Languages: Python, Bash, Shell Scripting
- Frameworks: PyTorch, TensorFlow, Scikit-Learn, Flask, Gradio, FastAPI, Streamlit
- Cloud & DevOps: AWS, Azure, GCP, Docker, MLflow, Azure DevOps
- ML Expertise: Generative AI, LLMs (Mistral, Gemma, LLAMA), Anomaly Detection, Federated Learning, CV/NLP
- Certifications: IBM | Microsoft | NVIDIA | AWS
- Offline Speech-to-Text + LLM Form Assistant: Transcription + Form Filling for NADRA
- LLMs on LAN: Deployed Gemma, Mistral, and LLAMA for internal users without internet dependency
- OCR Document Processing: Extracted data from passports, NICs using YOLO + Tesseract
- Federated Learning Frameworks: FL for anomaly & object detection (YOLOv5/v8)
- Biometric Background Removal: Deployed ICAO-compliant background replacer using MODNet
- Azure Functions for CEO Reporting: Auto-email work summaries from dashboards
- Reinforcement Learning for ECG Classification
arXiv:2401.04938 - Sepsis Stage Prediction via ML in ICU Patients
medRxiv:2022.03.15.22271655
- 85% uplift in NLP model accuracy using custom LLM fine-tuning
- Reduced cloud deployment time by 70% using Azure DevOps & CI/CD
- Improved fraud detection by 20% using ML at Askari Bank
- Managed 10+ production ML endpoints with 99.9% uptime
🧠 Let’s collaborate on anything AI — from LLMs to real-time data intelligence.