Data Scientist & ML Engineer _

Aryan
Garg

MSc Data Science · Napoli Federico II · ● open_to_work

I build machine learning systems that make predictions people trust — from end-to-end pipelines and deep learning models to deployed Streamlit applications.

view_projects() → GitHub ↗ email_me ↗

XGBoost TensorFlow Streamlit Scikit-learn Kaggle VGG16

25+

Engineered features
IPL prediction

83.7%

Cross-validation
accuracy · Kaggle

ML algorithms
benchmarked

Vehicle classes
classified · VGG16

◆ Python ◆ XGBoost ◆ LightGBM ◆ CatBoost ◆ TensorFlow ◆ Keras ◆ VGG16 ◆ Transfer Learning ◆ Scikit-learn ◆ Pandas ◆ NumPy ◆ Streamlit ◆ Feature Engineering ◆ Cross-Validation ◆ Deep Learning ◆ CNN ◆ EDA ◆ SQL ◆ Git & GitHub ◆ Python ◆ XGBoost ◆ LightGBM ◆ CatBoost ◆ TensorFlow ◆ Keras ◆ VGG16 ◆ Transfer Learning ◆ Scikit-learn ◆ Pandas ◆ NumPy ◆ Streamlit ◆ Feature Engineering ◆ Cross-Validation ◆ Deep Learning ◆ CNN ◆ EDA ◆ SQL ◆ Git & GitHub

about_me.py

Data tells a story.
I build the pipeline to tell it.

I'm a Data Scientist and ML Engineer pursuing my Master's in Data Science at Università di Napoli Federico II, Italy. My work focuses on building end-to-end ML pipelines: feature engineering, model benchmarking, evaluation, and real deployment.

My projects span predictive modelling, deep learning, computer vision, and Kaggle competitions — all built with reproducibility, real-world deployment, and meaningful metrics in mind.

Originally from Jaipur, Rajasthan · Now building in Naples, Italy

projects shipped

83.7%

best kaggle cv score

25+

features engineered

years in data science

locationNaples, Italy · Jaipur, India

emailnaryagarg@gmail.com

phone+91 894 901 2489

githubAryan-garg-1 ↗

linkedinaryan-garg ↗

languagesEnglish, Hindi

statusopen to opportunities

focusML · Deep Learning · AI

tech_stack.json

Tools I work with.

// languages

Python
SQL
HTML & CSS

// machine learning

XGBoost · LightGBM · CatBoost
Scikit-learn · Random Forest
Gradient Boosting
Feature Engineering
Cross-Validation · HParam Tuning

// deep learning & cv

TensorFlow · Keras
CNN · Transfer Learning
VGG16 · Fine-tuning
Data Augmentation
EarlyStopping · ModelCheckpoint

// data & analysis

Pandas · NumPy
Matplotlib · Seaborn
EDA · Data Cleaning
Statistical Analysis

// deployment & tools

Streamlit
Google Colab (GPU)
Jupyter Notebook
Git · GitHub
Pickle · Model Serialization

// evaluation metrics

ROC-AUC · F1-Score
Confusion Matrix
Log Loss · Precision/Recall
Calibration Curve
Cross-Validation Score

projects.py

Things I've built.

★ featured

2024 – 2025

Machine Learning Deployed · Live

IPL Match Winner Prediction

// end-to-end ML system for predicting IPL match outcomes

Built a complete ML pipeline using historical IPL data (2008–2025). Engineered 25+ cricket-specific features covering batting strength, bowling strength, venue performance, team form, and squad-level aggregates. Benchmarked 6 algorithms (LR, RF, GBM, XGBoost, LightGBM, CatBoost) and deployed the best model as a live Streamlit app.

Model Training ↗ Live App ↗ App Repo ↗

PythonXGBoost LightGBMCatBoost Scikit-learnStreamlit PandasFeature Engineering

↗

2024

Deep Learning Computer Vision

Indian Vehicle Classification

// VGG16 transfer learning — 5 vehicle categories

Deep learning classifier for Auto, Bus, Tempo Traveller, Tractor, and Truck using VGG16 transfer learning. Two-stage training (feature extraction → fine-tuning), data augmentation, EarlyStopping, ModelCheckpoint. Evaluated with confusion matrix, accuracy/loss curves, prediction visualizations.

TensorFlowKeras VGG16Transfer Learning Google ColabNumPy

↗

2023 – 2024

Kaggle ML

Titanic Survival Prediction

// Kaggle competition — 83.7% CV accuracy

Binary classification with systematic feature engineering, missing value imputation, categorical encoding, and hyperparameter tuning. Achieved 83.7% cross-validation accuracy with XGBoost ensemble. Fully reproducible training, evaluation, and submission pipeline.

PythonXGBoost Scikit-learnPandas Feature EngineeringSeaborn

↗

2023

Full-stack

MacroServer + MacroClient

// Python backend + native macOS client

Python-powered backend server for macro automation, paired with a native macOS Swift client. Built from architecture to deployment — systems thinking across two languages.

// offline proximity chat via MultipeerConnectivity

iOS chat app using Apple's MultipeerConnectivity for offline communication over Bluetooth and Wi-Fi Direct. Custom SwiftUI bubbles, JSON messaging protocol, local data persistence.

SwiftSwiftUIMultipeerConnectivity

// focus app shipped on TestFlight

Productivity app with Pomodoro timers, emotion journaling after study sessions, and session history via Core Data. Iterated from real TestFlight user feedback.

SwiftSwiftUICore DataTestFlight

↗

education.json

Where I learned.

Nov 2021 – Present

MSc Data Science

Università di Napoli Federico II

▸ Naples, Italy

Machine LearningData Analysis Predictive ModelingData Processing

Sep 2022 – Jun 2023

Apple Developer Academy Program

Apple Developer Academy @ Federico II

▸ Naples, Italy

Swift & SwiftUIProduct Design Figma · MiroTeam Collaboration

Jan – Apr 2023

Federico II Apple Foundation Program

Università di Napoli Federico II

▸ Naples, Italy

Jul 2017 – Jul 2020

Bachelor of Computer Applications

S.S. Jain Subodh P.G. Autonomous College

▸ Jaipur, Rajasthan, India

Data StructuresAlgorithms Database ManagementWeb Dev

// chapter_02 · the_traveller

Beyond the data,
there is a world.

From the pink city of Jaipur to the streets of Naples — collecting places the way others collect data points. Stories, photographs, and routes from the road — coming soon.

[ coming_soon · stories_loading... ]

contact.sh