N M EMRAN HUSSAIN

Work Status: Work permit (EAD) active | Immediate availability | Open to relocation.

Business Data scientist with an MS in Business Analytics from George Washington University, School of Business. Experienced in building data pipelines, developing machine learning models, and bridging technical analytics with business strategy. Background in predictive modeling, statistical analysis, and extracting actionable insights from complex datasets.

LinkedIn GitHub Email Phone

Technical Skills

Programming & Core Libraries

Python (Pandas, NumPy, Scikit-learn, TensorFlow, Matplotlib), SQL, R (Tidyverse, ggplot2).

AI and Generative Systems

Generative AI, Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Transformers, LangChain, Vector Databases (ChromaDB), NLP (Word2Vec, FastText, Sentiment Analysis).

Tools & Cloud Platforms

GCP (Vertex AI, BigQuery ML), Gemini 2.5 Flash, Vertor Search (RAG), ChormaDB, SHAP, DuckDB, Streamlit, Docker, Cloud Run, LinUCB, XGBoost, LangChain, Spark (PySpark), PyTorch, PostgreSQL, Version Control, Git/GitHub.

Business Strategy Analytics

A/B Testing, CLV, CAC, Churn Prediction, ROI Analysis, ROMI Analysis, AARRR Funnel, Stakeholder Management.

Data Engineering & MLOps

ETL Pipelines, Web Scraping, API Integration, CI/CD for ML, Data Cleaning & Normalization.

Visualization & Reporting

Seaborn, Matplotlib, and Plotly.

Featured Projects

Cardiovascular Risk Assessment & Triage Assistant

Vertex AI BigQuery ML Vector Search (RAG) SHAP Streamlit Docker Cloud Run

Pipeline Architecture: Architected a 4-agent clinical pipeline utilizing Vertex AI Gemini to parse unstructured vitals and a BigQuery ML boosted tree to predict heart failure mortality with SHAP explainability.
RAG & Deployment: Engineered a RAG system via Vertex AI Vector Search for medical guideline retrieval, deploying the solution to Cloud Run as a serverless Streamlit dashboard with active MLOps drift monitoring.

View on GitHub

Diabetic Readmission Risk & Clinical BI Copilot

Groq API DuckDB XGBoost SHAP Plotly Hyperparameter-Tuning Python

Architecture: Architected a 3-agent clinical triage application leveraging the Groq API for rapid LLM orchestration to evaluate 30-day diabetic readmission risk across 101,700+ patient records.
Data & ML Pipeline: Automated DuckDB SQL data extraction and XGBoost feature predictions to drive clinical insights.
Visualization & Explainability: Pipelined executive BI summaries and SHAP risk drivers into interactive Plotly cohort visualizations directly within Google Colab.

View on GitHub

Integrated E-Commerce Analytics: Predictive Churn & Contextual RAG Recommender

Gemini 2.5 Flash Gemini-embedding-001 Reinforcement Learning (LinUCB) ChromaDB LangChain RAG

System Architecture: Developed an end-to-end decision-support system integrating predictive churn modeling with Retrieval-Augmented Generation (RAG) to provide qualitative context for personalized customer interventions.
Data Pipeline: Engineered an automated Python pipeline utilizing Gemini and ChromaDB to synthesize structured metrics from the UCI Online Retail dataset with unstructured support feedback and interaction logs.
Business Impact: Transformed raw historical data into context-aware retention intelligence, enabling stakeholders to deploy proactive strategies that drive measurable improvements in long-term customer retention.

View on GitHub

LeNet-5 CNN Digit Classification

CNN TensorFlow, Keras Sigmoid/tanh

Architecture: Developed a compact LeNet-5 Convolutional Neural Network (CNN) using TensorFlow and Keras.
Performance: Successfully trained and evaluated the model to achieve 98%+ test accuracy on the MNIST digit dataset.
Techniques: Showcased fundamental deep learning and model optimization skills using Sigmoid and tanh activation functions.

View on GitHub

Credit Risk Model Bias Mitigation

Scikit-Learn XAI (SHAP & EBMs) Responsible AI

Analysis: Evaluated baseline credit risk models for demographic bias and discriminatory feature interactions using Scikit-Learn and Explainable AI (SHAP, EBMs).
Mitigation: Applied pre-processing and post-processing strategies to re-balance decision thresholds and reduce disparate impact across historically marginalized groups.
Impact: Successfully improved equal opportunity and demographic parity metrics while maintaining overall predictive power and Responsible AI compliance.

View on GitHub

Capital Bikeshare Analytics

Ensemble Modeling Predictive Modeling Random Forest Gradient Boosting Neural Network Linear Regression Ridge Lasso Elastic Net Forecasting Ridership Optimization of Decision Cost

Data Pipeline: Engineered temporal and weather features to build an predictive analytics pipeline for forecasting daily ridership and fleet availability.
Machine Learning: Benchmarked diverse algorithms—including regularized regressions, Random Forest, Gradient Boosting, and Neural Networks—using cross-validation and hyperparameter tuning.
Business Impact: Leveraged ensemble approaches to effectively capture complex, non-linear demand patterns, supporting data-driven station rebalancing and inventory allocation.

View on GitHub

Unsupervised Spatial Analysis

Scipy & Matplotlib Within-cluster Sum of Squares (WSS) K-Medians & K-Medoids

Objective: To Explore Capital Bikeshare data by mapping station locations and applying unsupervised learning to visualize geolocation clusters and usage patterns.
Modeling Strategy: Applied the elbow method on within-sum-of-squares (WSS) plots to determine the optimal cluster count, successfully identifying K = 10 as the inflection point that balances clustering quality and diminishing returns.
Impact: This project delivers an optimized operational framework for bikeshare network rebalancing by benchmarking advanced clustering techniques (K-Means, K-Medians, K-Medoids, and Hierarchical).

View on GitHub

Regression Modeling with H2O For Loan Prediction: GLM VS ANN

H2O.ai GLM & ANN L1/L2 Regularization

Objective: Evaluated the critical trade-off between predictive accuracy and regulatory interpretability for financial loan default risk assessment.
Modeling Strategy: Trained and benchmarked Generalized Linear Models (GLMs) with L1/L2 regularization against complex Artificial Neural Networks (ANNs) using the H2O.ai platform.
Impact: Delivered a comparative framework proving that transparent linear models can maintain competitive predictive power against deep learning architectures, establishing an interpretable and ethical risk pipeline.

View on GitHub

IMDb Sentiment Classification via FastText & Autotune

FastText NLP Optimization Autotune

Objective: Addressed underfitting and noise in unstructured text classification to build a robust, scalable system for high-throughput customer feedback processing.
NLP Pipeline: Engineered and standardized a supervised natural language processing pipeline on 25,000 IMDb reviews using Python, FastText, and N-gram feature engineering.
Optimization & Impact: Systematically tuned model parameters utilizing Autotune, successfully boosting classification accuracy from a 50% baseline to 87.4%.

View on GitHub

Professional Experience

Business Data Scientist

July 2026 - Present

Community Dreams Foundation (Skills-Based Volunteer) (Project: Elpsan Bio)

Sebring, Florida (Remote)

Evaluated predictive modeling systems for an AI-driven drug repurposing (i.e. Covid-19) initiative that analyzes illness behavior and tests millions of 3D drug combinations.
Analyzed the strategic business viability of candidates utilizing the FDA 505(b)(2) regulatory pathway, demonstrating a proposal of 70% to 90% reduction in development costs while nearly tripling clinical approval success rates to approximately 30%.

Business Analyst

May 2025 - July 2025

Ascend Technology Inc. Full-time

San Jose, CA

Designed and automated KPI dashboards using SQL and Python, tracking user engagement and system operational metrics across cross-functional sprints.
Wrote optimized SQL queries against relational cloud databases to aggregate multi-table transaction logs, reducing weekly manual report generation time.
Partnered with product and engineering managers during weekly Agile ceremonies to translate business workflow bottlenecks into technical user stories and data requirements.

Data Analytics Lead

January 2019 - October 2023

Ampere Ltd.

Dhaka, Bangladesh

Built seasonal time-series forecasting models in Python (Pandas, NumPy) and R for LED lighting and electrical components, reducing inventory forecast variance across core product lines.
Engineered predictive logistics models in Python (SciPy) using Monte Carlo simulations to map freight transit, providing supply chain planners with actionable bottleneck alerts during Q4 peak seasons.
Automated financial data pipelines in Python to calculate dynamic unit landed costs—integrating volatile currency fluctuations and tax incidence—replacing manual extraction workflows and improving reporting consistency across finance teams.

Assistant Manager

June 2009 - December 2018

Anrhus & Shuzos Co.

Dhaka, Bangladesh

Applied statistical hypothesis testing and regression analysis using R (dplyr, ggplot2, tidymodels) and Python (Scikit-Learn), automating competitive benchmarking pipelines to model demand variations for industrial machinery.
Streamlined inventory analytics and localized data tracking workflows to optimize infrastructure logistics and improve audit reporting consistency.
Applied root-cause data analysis to manage technical infrastructure operations, replacing manual tracking with automated pipelines to support complex, multi-stakeholder operational projects.

Let's Get In Touch

I am currently open to discussing new opportunities in Business Data Science, Data Analyst, Market Analyst, and Applied AI Engineering job roles. Whether you have a question or just want to connect, my inbox is always open.

DOWNLOAD FULL RESUME

Available for full-time roles

Email Me

nmemran.hussain@gwu.edu

Call Me

(408) 620-0255

LinkedIn GitHub