Mahmoud Trigui

About Me

Bridging data precision
with business strategy

Data Scientist with over 10 years of experience, blending technical precision with strategic insight. Holding a Statistics and Data Analysis Engineering degree, I have built a career on turning complex, business-driven structured data into decisions that matter — combining deep mathematical foundations with iterative, hands-on exploration.

My passion lies in feature engineering: transforming raw data into high-impact representations that align models closely with real-world objectives. I believe the gap between a good model and a great one is almost always in the features, not the algorithm. I approach every project with a different perspective — asking what the data is really saying before deciding how to model it.

Active in competitive ML (Zindi, IEEE — 19 competitions), continuously testing ideas against real problems across diverse domains. Fluent in both R and Python ecosystems, with production experience on GCP, Dataiku, and BigQuery.

🇹🇳 Arabic — Native 🇬🇧 English — C2 Proficient (EFSET 75) 🇫🇷 French — Fluent

📡 Telecom — Mobile Customers 💼 VC & Investments 🌐 Internet Provider — Orange

Tech Stack

Tidyverse

Tidymodels

Shiny

Plumber

R Markdown

H2O

mlr3

Python

Scikit-Learn

XGBoost

LightGBM

MLForecast

Prophet

Jupyter

Toad

MySQL

BigQuery

GCP

Dataiku

SAS

LLM

API

JSON

Postman

IBM Watson

Power BI

Looker

Excel

Git

VS Code

Jira

Education

Built on strong foundations

2016

Data Science: Data to Insights

MIT xPRO — Online

Advanced online program from MIT covering the full data science pipeline — from raw data to actionable insights. Reinforced production-oriented thinking and rigorous statistical reasoning.

2015

Engineering Degree — Statistics & Data Analysis

ESSAI — Higher School of Statistics and Data Analysis, Tunisia

Engineering program combining statistics, probability, data mining, machine learning, and applied mathematics. Graduated with a solid foundation in both theory and practice.

2011

Preparatory Institute for Engineering Studies

IPEIS — Mathematics & Physics track, Tunisia

Intensive university-level preparation in Mathematics and Physics — the Tunisian equivalent of the French Classes Préparatoires. This program taught me how to decompose complex problems, think rigorously under pressure, and build the mental frameworks that still shape how I approach data science today.

2008

Baccalauréat — Computer Sciences

15 November High School, Tunisia

Graduated with First Class Honors — the starting point of a journey built on curiosity, precision, and a love for problem-solving.

What I Do

End-to-end ML expertise

From raw structured data to production-ready models — with business context at every step.

🤖

Predictive Modeling

Full-stack ML pipelines from data cleaning to production evaluation. Deep feature engineering combined with tree models, ensembles, and stacking — with proper validation and business-metric alignment.

#Feature Engineering#Tree Models#Ensembles #Stacking#Clustering#Recommendation System#Fraud

📈

Time Series Forecasting

Multi-horizon forecasts with semi-automatic parameter search, multi-seed robustness, and production-ready pipelines. Weekly, daily, and seasonal patterns across thousands of IDs.

#MLForecast#Deep Forecast#R Forecast#Decomposition

📊

Analytics & Strategy

Translating analytical findings into business decisions. Exploratory analysis, statistical inference, clustering, segmentation, and interactive dashboards from raw operational data.

#SQL#Statistics#Segmentation#CVM#WebApp

🔤

NLP & LLM Integration

Text classification, topic modeling, sentiment analysis, and LLM-augmented pipelines. Multilingual including Arabic. Structural NLP signals often outperform raw text sentiment.

#LDA#Sentiment#LLMs#APIs

Ideas & Insights

10 years of DS thinking,
one post at a time

Technical ideas from real projects — not tutorials, not theory. Each post reflects a decision made under real constraints.

Forecasting · Architecture

I tuned MLForecast itself before I tuned the model

In a weekly forecasting framework, the search space was larger than hyperparameters. Step 0 searched the meta-space first — lag mode, search strategy, seed logic, and validation method.

Read full post → LinkedIn →

Forecasting · Search Design

This wasn't hyperparameter tuning. It was forecasting architecture search.

The parameter space had architectural layers. Every choice changed the type of system being built, not just how well a fixed system performed.

Read full post →

Healthcare · Optimization

The bottleneck in emergency response is not the ambulance — it is the wait before dispatch

Predicting ambulance response time in Nairobi revealed what looked like a traffic problem was really a dispatch queue problem.

Read full post →

View All Insights →

Projects

Real problems, real constraints

Selected work from competitions and client engagements.

Semi-Automatic Forecasting Framework

Weekly multi-horizon forecasting for 4 KPI targets across thousands of IDs. Semi-automatic pipeline: Step 0 meta-search → lag search → feature selection → hyperparameter tuning. Multi-seed robustness, dead-ID routing, BigQuery integration, rolling backtest.

Telecom · Orange4 targets52-week horizonGCP · BigQuery

View on GitHub →

Nairobi Ambulance Response Time

Predicted ambulance response times in Nairobi using geospatial features, dispatch queue signals, and temporal patterns. Identified that the true bottleneck was pre-dispatch wait, not road distance.

HealthcareZindi CompetitionGeospatial FeaturesTop Ranking

View on GitHub →

IEEE Customer Support Escalation

Predicted support ticket escalation before it occurs using structural features: response latency, re-open rate, message length progression. Structural signals outperformed raw text sentiment by a wide margin.

NLPIEEE CompetitionEarly Warning SystemStructural NLP

View on GitHub →

View All Projects →

Experience

Where the data was real

Feb 2024 – Present · Sfax, Tunisia (Hybrid)

Data Scientist — Sofrecom (Orange Group)

Built a semi-automated 12-month and 52-week forecasting system for multiple KPI targets using MLForecast. Designed 3-step optimization (lag selection → feature identification → hyperparameter tuning). Delivered forecasts via PowerBI and Looker Studio dashboards.

Developed a PDF audit information extraction system using GPT-4 with multi-page processing, Pydantic validation, and automated JSON output. Built FTTH access orders forecast achieving ~5% RMSE. Migrated co-financing prediction model to Dataiku production workflows.

MLForecastGCPBigQueryGPT-4DataikuPowerBI

Feb 2021 – Jul 2025 · Remote

Data Science Mentor — Codementor

⭐ 5 stars rating from 59 reviews · 94 live sessions and projects. Mentored data scientists across all levels on ML, forecasting, feature engineering, R, and Python.

MentoringRPythonML⭐ 5.0 · 59 reviews

Mar 2021 – Feb 2023 · Spain (Remote)

Data Scientist — Kiota Intelligence (VC Industry)

Built predictive models for the VC industry entirely in R: startup survival prediction, Series A funding probability, pre-valuation modeling, and investor-startup matching. Developed Shiny dashboards, automated RMarkdown reports with CSS styling, and multidimensional outlier detection systems.

RShinyVCNLPPredictionFeature Engineering

2018 – 2021 · Remote

Freelance Data Scientist — Sintegra Labs

Freelance data science missions across multiple client projects. Machine Learning Training Development . NLP Classification System . IoT Sensor Analytics . Social Media Analytics Platform

FreelancePythonRMLNLP

Jan 2018 – Mar 2020 · Tunis, Tunisia

Data Scientist — Tunisia Telecom (CVM & Data Mining)

Customer segmentation using K-means on behavioral data. Multi-SIM user detection with custom scoring algorithm. Family community detection within large networks. Collaborated with SAS, KPMG, and Business&Decision experts on churn, cross-sell, and community link analysis.

SASSQLSegmentationClusteringTelecom

Jun 2016 – Dec 2017 · Tunis, Tunisia

Data Analyst — Tunisia Telecom (CVM & Data Mining)

SQL ad-hoc requests and dashboard development for CVM performance, network quality KPIs, customer segmentation profiles, and sales analytics. Implemented VBA automation for PowerPoint reporting. Designed targeted marketing campaigns and Try & Buy offer frameworks.

SQLSASVBACVMDashboards

Feb 2022 – Dec 2022 · Sfax, Tunisia

Part-time Statistics Lecturer — International School of Business

Taught statistics to first and third-year business students.

StatisticsTeaching

Aug 2019 – Jan 2021 · Remote

Competitive ML — Zindi & IEEE Platforms

19 competitions across healthcare, telecom, environmental, NLP, and financial domains. Notable results: 4th place COVID-19 spread prediction (884 competitors), Top 11% Uber Cape Town road incidents, 6th place AI Hackathon Tunisia 2019 (fraud detection), Top 13% Financial Inclusion in Africa.

ZindiIEEER19 Competitions

Jun 2015 – Dec 2015 · Toulouse, France

Research Statistician — LAAS-CNRS

Developed an automatic home security management system (Smartfox Project with MyFox). Built behavioral pattern recognition for homeowner identification, real-time intrusion detection, and automated alarm management. Published research book "Gestion Automatique d'un Système de Sécurisation des Biens à Domicile" (European University Editions).

ResearchStatisticsPattern RecognitionPublication

Beyond the Data

The person behind the models

I am a husband and a father of two — a son and a daughter. Family is where I recharge and find perspective. The time spent with them quietly shapes how I think about patience, long-term thinking, and what is truly worth building.

I love to travel and discover new landscapes — different cultures, different ways of seeing the world. Nature, in particular, has a way of resetting the mind. Mountains, coastlines, open skies — every landscape is a reminder that the most interesting patterns are not always in a dataset.

I stay active through walking and swimming — simple habits that keep the mind clear. And at home, our cats are a constant, calming presence that the whole family adores.

I believe that a balanced life makes a better data scientist: curiosity fed by experience, perspective sharpened by the world outside the screen.

👨‍👩‍👧‍👦 Family ✈️ Travel 🌿 Nature 🏊 Swimming 🚶 Walking 🐱 Cats 🌍 Discovery 📖 Reading 📚 Continuous Learning

INFJ-A

Assertive Advocate

Role Diplomat

Strategy Confident Individualism

Mind 72% Introverted

Nature 92% Feeling

Tactics 92% Judging

Rare combination of analytical depth and human empathy — which is exactly why I care about the business impact of every model I build.

Contact

Let's work
together

Available for consulting, freelance, contract, and full-time roles — remote, hybrid, or open to relocation. Comfortable working across multiple domains with a preference for data-rich, business-driven problems.

✉

trigui.mahmoud7@icloud.com trigui.mahmoud7@gmail.com trigui.mahmoud7@outlook.com

📱 +216 97 304 501 (WhatsApp)

🔗 linkedin.com/in/mahmoudtrigui

💻 github.com/mah-trigui

🎓 codementor.io/@mahmoudtrigui — ⭐ 5.0 · 59 reviews · 94 sessions

Available for new projects

Feature Engineering MLForecast Time Series Decomposition Forecasting LightGBM XGBoost Clustering Segmentation NLP LLMs Web App R Markdown SQL Oracle DB SAS-Guide SAS E-Miner Dataiku BigQuery GCP Python R CRISP-DM Hypothesis Testing ANOVA Data Analytics Dimensionality Reduction Recommendation System Network Analysis Geospace Analysis Spatial Data Embedding Sampling Techniques Decision Rules Data Storytelling CVM Churn Fraud Detection Sentiment Analysis Topic Modeling IBM Watson PowerBI Looker Studio VBA Statistical Learning Ensemble Modeling Stacking Cross-Validation Profiling ABT Construction Plumber Tidyverse Shiny Prophet Deep Learning Scikit-Learn JSON SAS Programming Git VS Code CSS Styling Automated Reporting Outlier Detection Temporal Clustering Startup Survival Pre-Valuation Modeling Investor Matching K-Means Decision Trees Data Science Predective Modeling SVM LDA Text Classification Facebook API IoT Sensors Weight Prediction Pattern Recognition Behavioral Analysis Real-Time Detection Community Detection Cell ID Analysis Rolling Backtest Pipeline Automation Data Quality Checks Data Reliability Specification Mapping Business Strategy Marketing Campaigns Try & Buy Frameworks KPI Dashboards Network Quality Sales Analytics Mentoring Statistics Lecturer Feature Engineering MLForecast Time Series Decomposition Forecasting LightGBM XGBoost Clustering Segmentation NLP LLMs Web App R Markdown SQL Oracle DB SAS-Guide SAS E-Miner Dataiku BigQuery GCP Python R CRISP-DM Hypothesis Testing ANOVA Data Analytics Dimensionality Reduction Recommendation System Network Analysis Geospace Analysis Spatial Data Embedding Sampling Techniques Decision Rules Data Storytelling CVM Churn Fraud Detection Sentiment Analysis Topic Modeling IBM Watson PowerBI Looker Studio VBA Statistical Learning Ensemble Modeling Stacking Cross-Validation Profiling ABT Construction Plumber Tidyverse Shiny Prophet Deep Learning Scikit-Learn JSON SAS Programming Git VS Code CSS Styling Automated Reporting Outlier Detection Temporal Clustering Startup Survival Pre-Valuation Modeling Investor Matching K-Means Decision Trees Data Science Predective Modeling SVM LDA Text Classification Facebook API IoT Sensors Weight Prediction Pattern Recognition Behavioral Analysis Real-Time Detection Community Detection Cell ID Analysis Rolling Backtest Pipeline Automation Data Quality Checks Data Reliability Specification Mapping Business Strategy Marketing Campaigns Try & Buy Frameworks KPI Dashboards Network Quality Sales Analytics Mentoring Statistics Lecturer

محمود الطريقي

Senior Data Scientist

Bridging data precisionwith business strategy

Built on strong foundations

Data Science: Data to Insights

Engineering Degree — Statistics & Data Analysis

Preparatory Institute for Engineering Studies

Baccalauréat — Computer Sciences

End-to-end ML expertise

Predictive Modeling

Time Series Forecasting

Analytics & Strategy

NLP & LLM Integration

10 years of DS thinking,one post at a time

I tuned MLForecast itself before I tuned the model

This wasn't hyperparameter tuning. It was forecasting architecture search.

The bottleneck in emergency response is not the ambulance — it is the wait before dispatch

Real problems, real constraints

Semi-Automatic Forecasting Framework

Nairobi Ambulance Response Time

IEEE Customer Support Escalation

Where the data was real

Data Scientist — Sofrecom (Orange Group)

Data Science Mentor — Codementor

Data Scientist — Kiota Intelligence (VC Industry)

Freelance Data Scientist — Sintegra Labs

Data Scientist — Tunisia Telecom (CVM & Data Mining)

Data Analyst — Tunisia Telecom (CVM & Data Mining)

Part-time Statistics Lecturer — International School of Business

Competitive ML — Zindi & IEEE Platforms

Research Statistician — LAAS-CNRS

The person behind the models

Let's worktogether

Bridging data precision
with business strategy

10 years of DS thinking,
one post at a time

Let's work
together