Python Developer · Applied AI Builder · Quant ML / ML Finance Research-Oriented Profile

I build applied projects with Python, Machine Learning, SaaS, automation and quantitative research methodology.

My current direction is focused on:

• Quant ML Research;
• ML Finance;
• Python-based research pipelines;
• systematic trading research;
• realistic backtesting;
• statistical validation;
• applied AI products;
• SaaS and automation.

I am building my profile through real projects, continuous learning, certifications and disciplined technical practice.

I am a Python / Machine Learning developer building toward Quant ML Research, ML Finance, Research Engineering and systematic trading research.

I am not interested in isolated notebooks or superficial demos.
I like building complete systems, testing them seriously, understanding their limitations and improving them step by step.

My strongest work combines two directions:

1. Applied AI / SaaS product building
   Through Brixa, a real AI-assisted document management SaaS for construction companies.

2. Quant ML research
   Through research-grade financial ML pipelines with feature engineering, walk-forward validation, realistic backtesting, risk management, statistical testing and honest limitations.

I am especially interested in junior opportunities around:

• Quant ML Research Junior;
• Quant Research Junior oriented toward Machine Learning;
• Quant Developer ML;
• Research Engineer Finance;
• ML Engineer Finance;
• systematic trading research support;
• Quant Analyst / Data Scientist Finance;
• applied AI / ML Engineering roles.

I am open to banks, asset managers, hedge funds, prop trading firms, fintech companies and AI/product-driven startups where Python, Machine Learning, research discipline and product execution matter.

My current portfolio is structured around a clear progression:

Brixa:
  Applied AI SaaS / product engineering.

Project 1:
  Nasdaq Quant ML Research
  Single-asset directional Quant ML research on Nasdaq / QQQ.

Project 2:
  Market-Neutral Quant ML Alpha Research
  Cross-sectional residual alpha research, ETF ranking, portfolio construction and risk diagnostics.

Project 2 V1:
  Regime-aware portfolio overlay using ex-ante regime information.

Project 3:
  C++ quantitative research infrastructure, such as a backtesting engine, portfolio optimizer or risk engine connected to Python.

The long-term goal is to build a coherent profile across:

• applied AI;
• quantitative research;
• financial machine learning;
• statistical validation;
• portfolio construction;
• research engineering;
• Python and future C++ infrastructure.

Repository: market-neutral-quant-ml-alpha-research

This is my most advanced Quant ML research project so far.

It is a research-grade pipeline for cross-sectional residual alpha discovery on a liquid ETF universe.

The project includes:

• ETF universe construction;
• daily OHLCV data pipeline;
• data cleaning and quality checks;
• residual return modeling versus SPY and QQQ;
• residual momentum and reversal baselines;
• cross-sectional target engineering;
• Linear / Ridge / Lasso / Logistic baselines;
• Random Forest / Gradient Boosting / XGBoost comparison;
• regime/context feature engineering;
• feature ablation;
• controlled model selection;
• ML ranking diagnostics;
• score-to-portfolio construction;
• beta control;
• long-only, reduced-short and dollar-neutral candidates;
• transaction costs and short-cost stress tests;
• walk-forward / out-of-sample validation;
• advanced statistical tests;
• bootstrap and block bootstrap;
• regime and crisis analysis;
• ex-ante regime analysis;
• PCA / covariance risk diagnostics;
• final champion / challenger decision.

The project started with a market-neutral residual alpha hypothesis.

The strict market-neutral implementation was not validated as the final champion.

The final V0 conclusion is more nuanced:

The ML ranking signal appears most exploitable on the long side.
Long/short variants remain useful challengers, but they are more fragile.
The value of the project is the full research process: hypothesis, testing, portfolio construction, OOS validation, statistical analysis, regime diagnostics and honest conclusion.

This project demonstrates how I approach quantitative research: not by over-selling a backtest, but by testing a hypothesis seriously and accepting the results.

Repository: nasdaq-quant-ml-research

This project focuses on a long-only Nasdaq / QQQ directional ML strategy.

The goal was not to present a toy trading notebook.
The goal was to build a serious research pipeline that tests whether a machine learning model can extract a statistically defensible signal from Nasdaq / QQQ daily market data.

The project includes:

• QQQ / Nasdaq daily OHLCV data pipeline;
• financial feature engineering;
• relative features instead of raw price levels;
• target engineering using future 5-day returns;
• XGBoost classification;
• model benchmarking;
• feature ablation;
• feature importance stability;
• feature pruning;
• temporal train / validation / test split;
• yearly walk-forward validation;
• validation-based threshold selection;
• realistic next-open execution;
• fees and slippage;
• ATR stop-loss and take-profit;
• maximum holding period;
• gap-aware execution logic;
• risk-based position sizing;
• regime analysis;
• probability diagnostics;
• probability calibration;
• stress testing;
• trade-level bootstrap;
• hypothesis testing;
• confidence intervals;
• equity curve and drawdown analysis;
• honest limitations and roadmap.

The current strategy is not presented as a production trading system or a market-neutral alpha strategy.

It is better described as:

A long-only, risk-managed Nasdaq machine learning strategy with reduced market beta, realistic execution assumptions, walk-forward out-of-sample validation and statistically positive trade-level expectancy.

The edge is positive and statistically defensible, but still moderate.

The main strength of this project is the research methodology:

• leakage-aware validation;
• realistic backtesting;
• cost and slippage modeling;
• feature selection discipline;
• model comparison;
• bootstrap analysis;
• statistical interpretation;
• clear understanding of limitations.

Website: brixa.fr
Public showcase: brixa-public-showcase

Brixa is an AI-assisted document management SaaS for construction companies.

The production source code is private, but the public repository presents the product, architecture, screenshots and technical overview.

Brixa includes:

• Python / FastAPI backend;
• SaaS architecture;
• authentication;
• email confirmation;
• PostgreSQL;
• company workspaces;
• secure document upload;
• OCR;
• document classification;
• field extraction;
• business rules;
• human validation workflow;
• document library;
• responsive frontend;
• admin dashboard;
• Stripe Checkout;
• Stripe webhooks;
• free trial and subscription logic;
• Render deployment;
• SEO foundations with robots.txt and sitemap.xml.

This project required responsibilities across backend development, applied AI, document processing, SaaS architecture, product thinking, workflow UX, QA/testing and technical coordination.

I designed and developed Brixa end to end: Python/FastAPI backend architecture, document AI pipeline, OCR, business workflows, SaaS deployment, Stripe integration and iterative testing with construction-domain logic.

Repository: income-classification-ml-pipeline

A complete supervised Machine Learning pipeline on the Adult Income dataset.

It includes:

• preprocessing;
• feature engineering;
• imbalanced classification handling;
• Logistic Regression / Random Forest / XGBoost comparison;
• GridSearchCV;
• ROC AUC / PR AUC evaluation;
• saved ML pipeline;
• FastAPI prediction endpoint;
• Streamlit demo app.

This project shows a classical supervised ML workflow with API and interface packaging.

Repository: ml-breast-cancer-diagnosis

A supervised Machine Learning pipeline using scikit-learn’s Breast Cancer dataset.

It includes:

• preprocessing;
• model comparison;
• GridSearchCV;
• evaluation;
• model persistence.

Repository: ml-fruit-classification

My first applied Machine Learning pipeline project.

It uses a KNN classifier to classify fruits using tabular features such as mass, width, height and color score.

I study quantitative mathematics regularly to strengthen my ability to reason about ML finance, trading research, model validation and portfolio risk.

Current topics include:

• expectation;
• conditional probabilities;
• normal distribution;
• binomial distribution;
• Bayes theorem;
• evaluation of predictive models.

• mean estimation;
• variance and standard deviation;
• hypothesis testing;
• confidence intervals;
• correlation analysis;
• regression;
• trading strategy validation;
• statistical significance;
• robustness;
• bootstrap;
• out-of-sample validation;
• sensitivity analysis;
• alpha / beta analysis.

• linear algebra;
• vectors and matrices;
• matrix-vector products;
• dot products;
• transposes;
• covariance matrices;
• eigenvalues and eigenvectors;
• vector spaces;
• change of basis;
• NumPy matrix computations;
• differential and integral calculus;
• optimization;
• objective functions;
• parameters;
• constraints;
• overfitting and over-optimization;
• mathematical machine learning;
• loss functions;
• gradients;
• regularization;
• bias / variance;
• time series;
• stochastic processes;
• Brownian motion;
• Markov processes;
• Poisson processes;
• differential equations;
• conservation laws.

The goal is not only to code models, but to understand the statistical and mathematical logic behind model validation, signal robustness and risk.

• Machine Learning Specialization — DeepLearning.AI / Stanford Online
• Supervised Machine Learning: Regression and Classification — DeepLearning.AI / Stanford Online
• Advanced Learning Algorithms — DeepLearning.AI / Stanford Online
• Unsupervised Learning, Recommenders, Reinforcement Learning — DeepLearning.AI / Stanford Online
• Structuring Machine Learning Projects — DeepLearning.AI
• AI For Everyone — DeepLearning.AI
• Applied Machine Learning in Python — University of Michigan

• Scientific Computing with Python — freeCodeCamp
• Responsive Web Design — freeCodeCamp

• Python
• SQL
• JavaScript
• HTML / CSS

• pandas
• NumPy
• scikit-learn
• XGBoost
• feature engineering
• model benchmarking
• preprocessing pipelines
• classification
• probability calibration
• GridSearchCV
• time-series-aware validation
• model evaluation
• model persistence

• financial feature engineering
• residual return modeling
• cross-sectional ranking
• temporal split
• walk-forward validation
• out-of-sample validation
• realistic backtesting
• transaction cost modeling
• slippage assumptions
• beta control
• portfolio construction
• risk management
• Sharpe / drawdown / profit factor
• expectancy analysis
• bootstrap confidence intervals
• hypothesis testing
• feature ablation and pruning
• regime analysis
• PCA / covariance diagnostics

• FastAPI
• REST APIs
• PostgreSQL
• SQLAlchemy
• authentication
• protected routes
• Stripe Checkout
• Stripe webhooks
• Render deployment

• OCR
• document processing
• classification pipelines
• business rules
• human-in-the-loop validation
• workflow automation

• Git / GitHub
• VS Code
• Streamlit
• Jupyter Notebook
• Render
• Google Search Console

I work every day to improve my technical level.

My current learning priorities are:

• stronger Python autonomy;
• cleaner research code;
• deeper statistics and probability;
• linear algebra for ML and portfolio risk;
• optimization for portfolio construction;
• time series and stochastic processes;
• technical English;
• future C++ quantitative research infrastructure.

I use projects as my main learning engine: each project must produce code, results, interpretation, limitations and a better next question.

My target is to become useful in teams working on:

• Quant ML Research;
• systematic trading research;
• financial machine learning;
• research engineering;
• portfolio construction;
• applied AI systems.

I am ambitious, but I want my work to speak through concrete projects, honest methodology and continuous improvement.

• GitHub: github.com/remy-rm511
• Brixa: brixa.fr