Projects

Featured case studies, ordered by relevance to analytics, data engineering, and applied ML/AI roles.

• NBA Win Probability Engine

  A machine learning dashboard predicting NBA game win probability before tip-off and live, play-by-play, validated through a real-time 2025-26 Finals run.

  #machine-learning#sports-analytics#cloud#deployed#python

• WC2026 Format Evaluation

  A Snowflake data pipeline and Power BI dashboard testing whether FIFA's 48-team World Cup expansion hurt competitive balance, backed by five statistically validated analytics marts.

  #data-engineering#cloud#bi#sports-analytics#python#sql

• Levi's RAG — AI Due Diligence Copilot

  A retrieval-augmented generation tool that answers financial due-diligence questions about Levi Strauss & Co. strictly from its SEC filings, with every answer tiered by evidence strength.

  #llm#rag#deployed#python#nlp

• Intelligent AI Workflow Automation

  An operational decision-support system that predicts business-workflow risk from structured process data and pairs it with a fully separate, deterministic rules engine for KPIs, SLA alerts, and recommendations, surfaced in a live Streamlit dashboard.

  #machine-learning#data-engineering#deployed#python

• Supply Chain Disruption Analytics

  An end-to-end Azure pipeline that takes shipment-delay monitoring from raw CSVs to a statistically validated, predictive delay-risk scoring system with a Power BI dashboard.

  #data-engineering#cloud#bi#machine-learning#python

• Drug Interaction Checker

  A command-line tool that checks a patient's medication list for drug-drug interactions and ranks them by severity, grounding every fact in a curated local dataset rather than an LLM's medical knowledge.

  #llm#deployed#python

• Property Listing Generator

  Turns a subject property, a handful of comps, and a listing style into a structured listing that leads with the property's genuine differentiators, decided by a deterministic comparison engine before any writing happens.

  #llm#deployed#python

• MovieLens Recommender System

  A full movie recommender system combining ALS collaborative filtering with a LightGBM hybrid ranking model, deployed as a Streamlit app.

  #machine-learning#recommender-systems#deployed#python

• InsightPilot: An AI Data Analyst Agent

  An agentic AI tool that turns raw CSVs into structured business insight reports, keeping every numeric claim grounded in Pandas-computed statistics rather than LLM guesswork.

  #llm#ai-agent#machine-learning#deployed#python

• Letterboxd Movie Sentiment API

  A sentiment system for messy, sarcastic Letterboxd reviews — weak labels from star ratings, a TF-IDF baseline, a fine-tuned DistilBERT model, and a FastAPI backend.

  #nlp#machine-learning#python#deployed

• Sim2Real User Engagement Analysis

  A comparison of user-engagement and churn modeling between simulated anime-viewing data and real Steam gameplay data, examining how data realities shape modeling assumptions.

  #machine-learning#data-science#python

• Predicting Smartphone Addiction (Kaggle)

  A gradient-boosting pipeline predicting smartphone addiction risk from self-reported usage data for a Kaggle Playground Series competition, built around a validated cross-validation harness and a frozen-control feature-engineering process.

  #machine-learning#data-science#python