Data Science, Applied AI/ML & Cloud Infrastructure (MLOps)
BS Statistics (Data Science) @ University of the Punjab · Lahore, Pakistan
Architecting statistical data pipelines, leakage-free machine learning models, and reproducible cloud deployments on AWS and Linux.
| Domain | Technologies & Libraries | Engineering Standard |
|---|---|---|
| Data Science & Statistical Analytics | Python Polars Pandas NumPy SciPy Statsmodels PostgreSQL |
Parametric/non-parametric inference, MCAR/MAR missingness audits, IQR outlier quantification, SARGable analytical SQL |
| Applied Machine Learning | Scikit-Learn LightGBM CatBoost XGBoost Optuna |
Out-of-Time (OOT) & GroupKFold validation, cost-matrix threshold tuning, PR-AUC evaluation, Brier score calibration |
| Deep Learning & GenAI | PyTorch Google ADK Gemini 2.5 Hugging Face |
Multi-agent state machines, strict Pydantic schema contracts, hybrid BM25 + vector retrieval, @torch.inference_mode() |
| DevOps & MLOps Infrastructure | AWS (VPC, EC2, AMI, IAM) Docker Linux GitHub Actions FastAPI |
Containerized model serving, Staging/Prod VPC isolation, automated CI/CD pipelines, reproducible Linux builds |
- DataSnap — Automated statistical EDA engine auditing missingness structures (MCAR/MAR/MNAR), skewness/IQR outlier volume, and target-conditioned distributions (
Python,Polars,Pandas,Seaborn). - FraudLens — Financial transaction investigation and data drift monitoring system using Out-of-Time (OOT) validation and custom false-positive/false-negative cost matrices (
Python,LightGBM,FastAPI). - DocuAudit AI — Multi-page document intelligence pipeline with structured JSON schema validation and real-time token streaming (
Gemini 2.5,Python,TypeScript). - AWS Production Infrastructure — Custom AWS VPC networking, isolated Staging and Production EC2 environments, custom AMI builds, and PostgreSQL administration (
AWS,Linux,Bash).