― My background

I’m a Data Scientist who takes AI from notebooks to production. I build end-to-end solutions: ETL/ELT, ML/DL modeling, rigorous evaluation, and lightweight MLOps, always with security and governance in mind. I turn large, messy datasets into clear, deployable products and communicate results with crisp dashboards and documentation. I thrive in cross-functional teams and focus on impact: better decisions, lower risk, and visible ROI.

― Skills

• Python (NumPy, pandas, scikit-learn, XGBoost, TensorFlow), SQL, Git

• EDA & feature engineering; supervised/unsupervised; time series

• Model evaluation & tuning (ROC-AUC, F1, RMSE, MAPE; Grid/Random/Bayes)

• ETL/ELT & orchestration (Airflow/Prefect), dbt, Parquet

• APIs & deployment (FastAPI/Flask, Docker, PostgreSQL)

• Data visualization & dashboards (Plotly, Power BI, Tableau)

• Security, privacy & explainability (governance, SHAP/LIME)

• Product mindset: problem framing, KPIs, ROI, stakeholder communication

― Early Detection of Autoimmune Diseases

Predictive analytics for earlier referral and better outcomes.

• Built an end-to-end pipeline from clinical/lab data to risk scoring.

• Feature engineering, model selection/tuning, and robust evaluation; model explainability with SHAP to support clinicians.

• Exposed results via API + dashboard for monitoring and audits.
  Stack: Python (pandas, scikit-learn/XGBoost), SQL, Plotly, FastAPI, Docker.

• Designed experiments and metrics (task quality, recall, dependency, time-on-task).

• NLP analysis and statistical modeling to track effects over repeated AI usage.

• Interventions: prompting playbooks, spaced retrieval, and guidance to restore independence.
  Stack: Python, transformers, notebooks, Streamlit/Flask, privacy-preserving analytics.

― Measuring & Mitigating Cognitive Debt from AI Use

A human-in-the-loop framework to quantify and reduce AI-induced cognitive load.

Featured Projects:

― Crowd Density Estimation via Computer Vision

Real-time people density for safer, smarter spaces.

• Curated/annotated video datasets and trained CNN-based density models.

• Produced heatmaps and threshold alerts; optimized inference for edge/server.

• Delivered an API for integration with dashboards and alerting systems.
  Stack: Python, PyTorch/TensorFlow, OpenCV, YOLO/DeepSort, Flask API.