Nabil Shehadeh

Data Scientist

Résumé

Today

I'm a data scientist finishing an MS in Applied Data Science at USC. By day I'm a data engineer at T-CETRA, building AWS pipelines that clean, validate and flag anomalies before data reaches a model.

I'm also building KSHF Clear, which decides whether an aircraft part is real before it goes into stock. I like models whose answers have to hold up in front of an inspector, a clinician or an auditor.

KSHF Clear Current work

Every part, proven real.

Clear checks the paperwork, history and supplier behind every part at the receiving bench, confirms the certificate with the organisation that issued it, and seals the decision so anyone downstream can verify it.

See a walkthrough

  1. The paperwork passed.

    Between 2019 and 2023, AOG Technics sold more than 60,000 suspect engine parts on forged release certificates. Every one looked fine at receiving. Clear starts with what needs attention across every source.

    Clear home screen listing parts waiting at receiving, a suspended supplier and an unlisted approval

  2. Ten checks, each with a reason.

    The issuer's approval, the signatory, the part's history back to birth, your own stores and the wider network. Every result shows its evidence and the rule behind it, so an inspector can stand behind the call.

    Receiving view with checks run on a fuel pump and a recommendation to accept

  3. A score you can audit.

    Each check returns a weighted reason, and the weights add up to a 0 to 100 risk score: 20 or more quarantines the part, 50 or more rejects it. Suppliers are linked in a graph by shared address, phone, domain, director or bank. Every signed decision is kept as a labelled outcome, so the next step is learning those weights from real verdicts instead of setting them by hand.

    Clear rules view listing the ten versioned checks and the score thresholds

  4. Exposure in seconds.

    Start from a suspended supplier and Clear follows every link to certificates, parts, aircraft and operators. After AOG Technics, airlines spent months doing this by hand.

    Exposure view tracing a suspended supplier to linked parts and aircraft

  5. Answers you can check.

    Ask in plain English. Clear shows the plan it ran and the source snapshots it read, so nobody has to trust a black box.

    Ask Clear answering which aircraft carry parts from a suspended supplier, with its plan shown

  6. A record anyone can verify.

    Each decision is signed, hash-chained and sealed. The format is published as KSHF Proof, an open standard, so the whole industry can raise its bar and not only our customers.

    KSHF website headline: every part checked, every certificate confirmed

Click the screen to try it live.

Projects

Fraud Detection System

A six-model stacking ensemble on 285K anonymised card transactions, where only 0.17% are fraud. Served by FastAPI in under 100ms, with every score explained by SHAP.

PR-AUC
0.871
ROC-AUC
0.951
Recall@100
0.72
Latency
<100ms

Judged on PR-AUC rather than accuracy, because at 0.17% fraud a model that flags nothing is 99.8% accurate. Class weights and SMOTE handle the imbalance, and 50+ behavioural and temporal features feed the models.

Full analysis Source

PR-AUC by model Held-out test set. Higher is better.
Held-out test results by model
Model PR-AUC ROC-AUC Recall@100
Stacking ensemble 0.871 0.951 0.72
LightGBM 0.856 0.943 0.67
XGBoost 0.841 0.938 0.64
Random forest 0.823 0.934 0.61
Logistic regression 0.742 0.891 0.45
Autoencoder 0.623 0.789 0.34

Experience

  1. Since Oct 2026

    Data Engineer at T-CETRA

    Migrating legacy databases into Amazon Redshift, with automated data-quality and reconciliation checks so migrated data is trustworthy from day one.

  2. Jun to Oct 2026

    AI Researcher Intern at GE HealthCare

    Shipped a production LLM agent that turns plain-English questions into SQL, cutting data-request turnaround by about 35%. Built PyTorch vision models for X-ray systems.

  3. Since Sep 2023

    Content Data Analyst at the NFL

    Designed A/B experiments on content formats that drove a 10% lift in engagement.

  4. 2023 to 2024

    Data Science Research Assistant at Rainier Clinical Research Center

    Built Python pipelines with automated QC that cut data errors by 20% across 500+ study visits.

Education

  1. Dec 2026

    MS, Applied Data Science at USC Viterbi

    4.0 GPA. Machine learning, deep learning, data mining.

  2. 2023

    BA, Public Health at the University of Washington

    4.0 GPA.

  3. Certified

    Professional Data Engineer from Google Cloud