Business Analytics & AI Professional

Building Intelligent, Responsible & Impactful Solutions

View GitHub

Featured by

The George Washington University for applied analytics and responsible AI work

About Me

Building trustworthy AI systems that combine machine learning, explainability, and real-world impact.

I'm a Machine Learning and Data Science professional passionate about building AI systems that are accurate, interpretable, and impactful. My work spans predictive modeling, Retrieval-Augmented Generation (RAG), large language models, explainable AI, and modern data engineering, with a focus on developing responsible AI systems that support transparent and trustworthy decision-making.

At Children's National Hospital, I contribute to clinical AI research by developing and evaluating a Retrieval-Augmented Generation (RAG) pipeline for differential diagnosis. The system integrates FAISS vector search, domain-specific medical embeddings, structured medical knowledge graphs, and locally deployed large language models through vLLM to improve clinical reasoning and retrieval quality. This work focuses on advancing reliable AI-assisted clinical decision support while exploring future multimodal capabilities.

Previously, at The George Washington University, I developed an award-winning AI decision-support system that combines Explainable Boosting Machines (EBM), SHAP, and Retrieval-Augmented Generation (RAG) to deliver transparent, analyst-facing fraud investigations. This experience strengthened my interest in building AI systems that are both high-performing and explainable.

My experience includes designing end-to-end machine learning solutions, building and evaluating RAG systems, developing interpretable and bias-aware models, forecasting business outcomes, and translating complex data into actionable insights across healthcare, finance, and operational analytics.

Core Expertise
  • Machine Learning & Predictive Analytics
  • Large Language Models (LLMs)
  • Retrieval-Augmented Generation (RAG)
  • Explainable AI (EBM & SHAP)
  • Healthcare AI & Clinical NLP
  • Responsible AI & Model Evaluation
Areas of Interest
  • Machine Learning Engineering
  • Applied AI
  • Healthcare AI
  • Human-Centered AI
  • Trustworthy & Responsible AI
  • Neuromorphic Computing

My journey into AI began after seeing the impact of data-driven decision-making while working at Coca-Cola and continued through founding The Squat Junkie, where I applied analytics, experimentation, and predictive modeling to grow a fitness venture. These experiences reinforced my belief that the most valuable AI systems combine technical excellence with measurable real-world impact.

Professional Experience

AI Engineer, Analytics & Strategy

Freedom Buckets • Remote
August 2026 – Present
  • Designing the AI decision architecture for a financial decision platform, translating financial, behavioral, goal, and historical data into personalized recommendations.
  • Developing the behavioral intelligence and recommendation framework, including behavioral variables, decision logic, intervention history, learning metadata, and the Bucket Allocation Engine.
  • Building evaluation frameworks, product metrics, and data strategies to measure recommendation quality, user behavior, and system performance as the product moves toward beta.

Research Trainee

Children's National Hospital • Washington, DC
February 2026 – Present
  • Developed and evaluated a medical Retrieval-Augmented Generation (RAG) benchmarking pipeline for 7,663 clinical question-answering tasks across five MIRAGE benchmarks, comparing baseline LLM reasoning with retrieval-enhanced generation under reproducible evaluation settings.
  • Diagnosed retrieval and inference bottlenecks by upgrading the underlying model from Qwen2.5-3B to Qwen3-8B, improving overall benchmark accuracy from 46.8% to 58.7% (+11.9 percentage points), including +19.5 on MedQA-US and +16.6 on MMLU-Med.
  • Engineered reproducible preprocessing, retrieval, crash recovery, and automated evaluation pipelines to standardize large-scale medical AI benchmarking and support reliable model evaluation.

Applied Analytics Consultant

The George Washington University • Washington, DC
August 2025 – December 2025
  • Led an end-to-end fraud analytics consulting engagement, partnering with industry stakeholders to design an explainable AI decision-support system for post-alert fraud investigations.
  • Developed an explainable fraud analytics platform using Explainable Boosting Machines (EBM), SHAP, Retrieval-Augmented Generation (RAG), and Streamlit, reducing analyst investigation time by 30–50% and delivering an estimated $1M savings per 100M transactions.

Student Athletic Trainer (Performance Analyst)

The George Washington University • Washington, DC
October 2024 – December 2025
  • Developed real-time performance analytics dashboards using Python, SQL, and Tableau to monitor 50+ athletes and support evidence-based health and performance decisions.
  • Streamlined data collection and reporting workflows, reducing weekly reporting effort by 40% while improving data quality and reporting consistency.

Founder & Marketing Data Analyst

The Squat Junkie Fitness Studio • Lahore, Pakistan
March 2017 – August 2024
  • Founded and scaled a fitness business, building KPI dashboards that unified operational, financial, and customer data across 12+ business metrics to support strategic decision-making.
  • Developed customer segmentation and retention analytics that identified high-risk churn segments during COVID-19, improving operational efficiency by 20%.

Digital Media Marketing & Export Data Analyst

Mayfair Asian Food Industries Ltd. • Lahore, Pakistan
July 2017 – February 2018
  • Built real-time export, sales, and operational dashboards using Python, SQL, and Tableau, reducing reporting time by 50% and accelerating business decisions.
  • Developed pricing, inventory, and campaign performance analytics, increasing campaign ROI by 15%.

Sales & Commercial Data Analyst Intern

The Coca-Cola Company • Lahore, Pakistan
October 2016 – November 2016
  • Automated sales and distribution reporting by transforming regional sales data into Excel dashboards for commercial performance monitoring.
  • Validated and standardized regional sales datasets, improving reporting accuracy and supporting reliable operational decision-making.

Projects

FraudLens – From Alerts to Insights

Built an interpretable fraud-analysis assistant to help analysts understand why transactions are flagged. FraudLens combines a surrogate Explainable Boosting Machine (EBM), SHAP explanations, and a TF-IDF–based RAG knowledge layer to translate model behavior and rule-based fraud signals into clear, human-readable narratives for faster, auditable review workflows.

Read the technical article →

Python Streamlit EBM SHAP RAG TF-IDF NLP
Hospital++ — Explainable Medical Logistics Decision Support

Designed and deployed a full-stack explainable AI decision-support platform that transforms complex medical logistics data into real-time, interpretable operational insights. The system enables healthcare teams to proactively identify supply chain risks, inventory shortages, and operational bottlenecks through transparent machine learning predictions.

Built a microservice-based architecture using FastAPI, React, PostgreSQL, Docker, and NATS messaging, with Explainable Boosting Machines (EBM) providing interpretable risk scoring and feature-level explanations for every prediction.

View Live Demo →

Python FastAPI React PostgreSQL Docker NATS EBM Explainable AI Healthcare Analytics
Fair Lending – Bias-Aware Modeling

Developed interpretable models to evaluate and remediate bias in decision outcomes. Used EBM and SHAP with fairness metrics (AIR) to balance predictive performance with transparency and responsible model use.

Python EBM XGBoost SHAP Fairness Testing AIR
CECL Credit Risk & Macroeconomic Modeling

Built a CECL-aligned credit risk framework using macroeconomic indicators (GDP, unemployment, home prices, delinquency rates) to estimate lifetime expected credit losses. Applied ARIMAX models, scenario analysis, and backtesting to balance predictive accuracy with interpretability and regulatory alignment.

R ARIMAX Time Series Credit Risk CECL Stress Testing

Code and data not publicly available due to academic and data constraints.

Market Entry Analytics – Iowa Liquor Sales

Built an end-to-end SQL and Python pipeline on AWS with Power BI dashboards to identify demand patterns, underserved regions, and expansion opportunities using store performance and geographic insights.

PostgreSQL AWS Power BI ETL Data Modeling Market Analysis
Capital Bikeshare Demand Forecasting

Forecasted daily bike demand using trip and weather data to support planning and resource allocation. Applied PCA and regression models with cost-based evaluation; LASSO achieved RMSE = 4.40 and R² = 0.85.

Python scikit-learn PCA LASSO Ridge Forecasting

Technical Skills

Generative AI, LLMs & RAG
Retrieval-Augmented Generation (RAG) Large Language Models (LLMs) vLLM FAISS BioLORD PubMedBERT Knowledge Graphs NetworkX Clinical NLP Medical Embeddings MedGemma Qwen3 LangChain Prompt Engineering OpenAI API
Machine Learning & Explainable AI
Explainable Boosting Machines (EBM) SHAP PDP Regression Classification Clustering Random Forest XGBoost LightGBM SVM Forecasting ARIMA ARIMAX PCA Model Evaluation A/B Testing Fairness Testing Bias Mitigation
Software Engineering & Infrastructure
FastAPI React Docker Docker Compose NATS AWS Apache Spark ETL Pipelines Git GitHub Streamlit VS Code Jupyter Notebook
Programming & Data
Python SQL R PostgreSQL MySQL SparkSQL pandas NumPy SciPy scikit-learn statsmodels HTML/CSS
Analytics & Visualization
Power BI Tableau SAS Visual Analytics Matplotlib ggplot2 Excel Google Analytics

Working Style

Evidence-Driven

I validate ideas through experimentation, benchmarking, and measurable improvements rather than assumptions.

Trustworthy AI

I prioritize explainability, fairness, and transparency when developing machine learning and AI systems.

End-to-End Builder

I enjoy taking projects from research and modeling to APIs, user interfaces, and deployable applications.

Collaborative

I work closely with researchers, clinicians, and technical teams to build practical AI solutions for real-world problems.

Awards & Recognition

AI Case Competition Award

The George Washington University School of Business · Fall 2025

AI Case Competition Award Certificate - George Washington University

Awarded for an outstanding Business Analytics Practicum project focused on building an interpretable, analyst-facing AI system for decision support.

Certifications

Google Analytics Certification

Issued by Google | 2024

Mastered analytics fundamentals, audience segmentation, and reporting insights.

GoogleAnalyticsDigital Marketing
Data Analysis with R Programming

Issued by Google | 2024

Learned data manipulation, visualization, and statistical modeling using R.

RData AnalysisVisualization

Contact