Milton Mamani photo

Hi, I'm Milton Mamani.

Mathematical Physicist and final-year Master's student in Theoretical Physics at IFT UNESP. I apply my background in quantitative modeling and complex systems to build end‑to‑end data science solutions, with a focus on financial fraud detection and credit risk modeling using Python – from EDA and ML modeling to deploying interactive web apps (Streamlit) and automating workflows with n8n + Google Gemini. Proficient in Python, SQL, Power BI, and Google Cloud.

Learn about my journey

Education, technical skills & languages

From theoretical physics to data science: analytical rigor, modeling, and automation.

Academic Background

Bachelor’s Degree in Physics – Universidad Nacional del Altiplano (Peru)
M.Sc. candidate in Theoretical Physics – IFT UNESP (Brazil)
Academic Exchanges & Research Visits: UNI (Peru), UNSa (Argentina), UAM (Mexico), INAOE (Mexico), IFT UNESP (Brazil).
Thesis published in Classical and Quantum Gravity (IOP Publishing)

Recognition & Leadership

🏆 Graduated with Honors - Summa Cum Laude
🎓 Scholarships: Permanencia (4 years), PMESUT (English)
👥 President, CIEF Puno 2023 – Academic leadership and event organization [Event details]
🤝 Organizer, XLVIII Congresso Paulo Leal Ferreira (IFT) – International academic conference [Event details]

Tech Skills

Programming: Python (pandas, scikit-learn, Streamlit), SQL, Git/GitHub
Data Analysis & Visualization: Python, SQL, Power BI (DAX), Jupyter Notebooks, Exploratory Data Analysis (EDA)
Machine Learning: Classification models (Logistic Regression, Extra Trees, Random Forest, XGBoost), handling extreme imbalance, GridSearchCV, pipelines
Automation & Cloud: n8n, Google Cloud (BigQuery, Drive, Sheets, Gemini LLM), Telegram Bot API

Languages

Quechua: Native
Spanish: Native
Portuguese: Proficient (living in Brazil)
English: Proficient (Idiomas PUCP certified; Duolingo English Test certified; academic and international conference experience)

Certifications & Complementary Training

Complementary training in Data Science, Cloud Computing, and Artificial Intelligence.

Machine Learning using Python Certificate

Machine Learning using Python

Practical training in supervised and unsupervised learning using scikit-learn and pandas.

Databricks for Machine Learning Certificate

Databricks for Machine Learning

Foundations of collaborative notebooks, distributed data processing, and ML pipelines.

Google Cloud Platform Certificate

Introduction to Google Cloud Platform

Core concepts of cloud infrastructure, storage, compute, and data services.

Artificial Intelligence Training UNESP Certificate

Artificial Intelligence Training – UNESP

Applied AI concepts, ethics, and practical implementation projects.

Featured Projects

End‑to‑end data science & AI automation – every claim backed by code and metrics.

Core Quant Projects

Credit Risk

Credit Risk Predictor [EDA + Extra Trees, +10% accuracy]

Complete EDA + Extra Trees classifier with 64.8% accuracy, outperforming basic heuristics by ~10%. Includes probability outputs and live Streamlit demo. Notebook contains full exploratory analysis, feature engineering, and GridSearchCV.

Fraud Detection

Financial Fraud Detection [EDA + 94% recall, 0.3% fraud]

Deep EDA on 2.7M transactions + pipeline with 94% recall on highly imbalanced data (only 0.3% fraud). Deployed live app – catches 2,309 of 2,464 frauds in test. Notebook shows imbalance handling, scaling, and business rule analysis.

Analytics Pipeline

Forensic Fraud Analytics [BigQuery + Power BI]

Same dataset as the Fraud Detection project (6M transactions). Forensic SQL in BigQuery: identified money mule networks and loss spikes of $150M per hour. EDA compared with the Python notebook to validate findings. Interactive Power BI dashboard (DAX, self-joins).

Web App

AI Financial Solutions Hub [Multi‑Page Streamlit]

Centralized portal deploying both fraud and credit models. Real‑time predictions with confidence levels, caching, and persistent encoders.

Automation & Infrastructure

Automation

AI Accounting Agent [n8n + Google Gemini]

Telegram bot + Gemini LLM extracts invoice data (JSON) → Google Sheets. No OCR, pure multimodal AI. Full automation, zero manual entry.

Let's talk

I'm looking for opportunities in Data Science and quantitative analysis. Let's connect.

Find me on:

Milton y Melyssa

💖 Con todo mi amor, ésta página y éste proyecto está dedicado a Melyssa Thayss, mi esposa, mi mayor motivación, mi todo. Te amo de aquí a la luna a pasitos de tortuga. 💖