open to data analyst & data scientist roles

Wissam Ezzedine

Data analyst based in Madrid, doing my MSc in Business Analytics & Data Science at IE after a BCom in Business Technology Management (minor in Data Intelligence) at Concordia. I dig through a dataset until I find the one number that explains what's going on, then build something around it that a non-technical person can use.

4 deep-dives, 2 live

Case studies

livepythonstreamlitlightgbm
Bad-review rate by delivery performanceordinal, one hue, intensity = severity
80%53%27%0% 8.9%10.5%12.2%49.3%79.8% 7+ EARLY1-7 EARLYON TIME1-7 LATE7+ LATE

Olist Review Risk: what predicts a bad review on a Brazilian e-commerce marketplace?

Interactive Streamlit dashboard on 94k Olist orders across 6 relational tables. Ten sections: KPIs, delivery bars, category ranking, time trend, Brazil choropleth, LightGBM classifier with feature importance, ROC + confusion matrix, an interactive live-prediction widget, a two-proportion z-test, and VADER sentiment analysis on a 5,000-review sample of review text. Late delivery is the strongest signal I found: orders that arrive more than 7 days late get a 79.8% bad-review rate (1-2 stars), vs 8.9% for orders delivered 7+ days early.

Decision this informs For an Olist ops team, the gap between 79.8% and 8.9% points at delivery operations rather than products: late orders are 6.7% of orders but 32.5% of all bad reviews. The case for intervening is on orders that are running late before the review is written. My LightGBM model uses delivery delay as an input, so it scores orders after delivery; a pre-delivery version would need different features.
kagglepython
0-100 km/h acceleration by fuel typenominal, one hue per fuel type (Electric is the site blue), median and IQR printed under each box, outliers abridged
30s20s10s0s

Cars 2025: how 0-100 km/h acceleration differs by fuel type.

Exploratory analysis of 1,218 cars from a Kaggle 2025 car dataset, in Python (pandas, seaborn, matplotlib). I parsed the mixed-format spec fields with regex, then built histograms, barplots, a boxplot, violin and swarm plots, and a correlation heatmap. Electric cars have the fastest median 0-100 km/h time (about 5.2s) and diesel the slowest (11s).

What it suggests Fuel type looks like a useful first cut when comparing cars on 0-100 acceleration: the electric median is about 5.2s, petrol about 7s, diesel 11s. This notebook does not compare fuel with brand, and it does not analyse price, so it says nothing about pricing or margin.
sqlduckdbctes, window fns
Top 10 "hosts" by listing countordinal, one hue, brighter = more listings
100200300 UkioHomeclubMIT HouseBemadridFelipeDlux RoomsRaul Sanchez V.Match4roomDobo HomesDuke Host 388384345240215164156126122120

Madrid Airbnb: who runs the listings?

SQL deep-dive on 23k listings and 8.4M calendar rows in DuckDB (Inside Airbnb, March-April 2026 snapshot). Chained CTEs, window functions (NTILE, RANK), joins, conditional aggregations. The top 10% of hosts list 53.3% of all Madrid Airbnb listings. Two commercial operators, Ukio (388 listings) and Homeclub (384), list 770+ combined.

Decision this informs For Madrid's housing regulator, this snapshot shows supply concentrated in a small operator tier. 960 hosts (10%) list 53.3% of the 23,002 listings. Ukio alone lists 388, more than all single-listing hosts combined in 125 of Madrid's 128 barrios, and every one of its listings requires a stay of 30 nights or more. A regulator looking at this market would need to look at the operator tier as well as the individual listing.
livetableaupythonrest api
Top 6 teams by points, top 6 scorers by goals, all five leagues, 2025/26ordinal, one hue, intensity = rank
Top teams, points Barcelona Bayern Inter Real Madrid Arsenal Man City 94 89 87 86 85 78 Top scorers, goals Kane Haaland Mbappé Muriqi Thiago Lepaul 36 27 25 23 22 21
Barcelona (94) led on raw points; the top six span 16 points.Kane's 36 was the most in the five leagues, 9 more than Haaland. No ties among the six shown. Scorer counts are the top 20 per league from the API.Points are raw season totals: Bundesliga and Ligue 1 teams played 34 matches, the other three leagues 38. Bayern's 89 is 2.62 per game, the highest of any team. Source: Football-Data.org free tier, goals only (no xG).

Europe's Top 5 Leagues: how home win rates differ by league, and which teams earned more points than their goal difference.

Interactive Tableau dashboard comparing the Premier League, La Liga, Bundesliga, Serie A and Ligue 1 across the 2025/26 season. Home win rate is home wins over all finished matches (771 of 1,751 in total; one Ligue 1 fixture is missing from the API's finished matches). Pulled the data from the Football-Data.org REST API in Python, reshaped it with pandas into a team-match model, and built five pages: season snapshot, competitive landscape, season story, scouting signals, and top scorers with full tables.

Decision this informs Home win rate ran from 49% in La Liga (186 of 380) to 39% in Serie A (148 of 380) (2025/26, all finished matches). Across all five leagues the differences are not statistically significant at the 5% level (chi-square p = 0.07), so the gap is suggestive rather than proven, but it is a reason to check home advantage by league before assuming one value in an outcome or odds model. Teams above the points-vs-goal-difference line earned more points than a straight-line fit of points on goal difference (fitted per league) predicts. The largest in each league: Napoli +8.5, Aston Villa +8.0, RB Leipzig +7.0, Rennes +5.7, Athletic Club +3.6 points. One season cannot say whether such gaps are luck or repeatable, so I read them as teams worth a closer look, not as a forecast.

Coursework

DataCamp coursework log: graded, not case studies

A free DataCamp account is required to view these.

3 internships

Experience

Project Management Intern

Toters Delivery
Jul - Aug 2024
+9%promo sales over 6 weeks
−25%manual checking time
35tickets cleared, 1-day avg
  • Ran the digital voucher project end-to-end: scoped the work, assigned owners, launched 6 weekly promos on schedule.
  • Coordinated Marketing, Support and Operations to set offers and fix issues, clearing 35 support tickets with a 1-day average turnaround.
  • Built a competitor tracker and a weekly promo report; reduced manual checking and helped lift promo sales over six weeks.

Digital Marketing Analyst

Workden
Nov 2022 - Nov 2023
+21%engagement YoY
2.1% → 3.0%CTR over 10 weeks
34campaigns shipped
  • Planned and shipped 34 campaigns across Instagram, TikTok, email and web; lifted engagement 21% year over year.
  • Managed the content calendar (avg. 3 posts/week) and weekly reporting; raised click-through from 2.1% to 3.0% in 10 weeks and held it above 2.5% for the rest of the year.

Data Analysis Intern

CMA-CGM Group
Jun - Aug 2022
−40%report prep time
6% → 3%error rate
10dashboards built
  • Consolidated data from 5 sources into a standard model and built 10 dashboards, cutting weekly report prep time by 40%.
  • Automated 2 recurring Excel workflows and wrote SOPs; cut the manual error rate roughly in half.

Skills

languages
Python, SQL, Java, C#, SAS
python libraries
pandas, NumPy, SciPy
machine learning
scikit-learn, LightGBM, Classification, Feature importance, VADER (sentiment), Linear Regression, Logistic Regression, Clustering, PCA
visualization & bi
matplotlib, seaborn, Plotly, Tableau, Power BI, Qlik, Excel, Lucidchart
tools & data sourcing
Jupyter, DuckDB, Streamlit, REST APIs
spoken3 native, 1 beginner
Français (native), English (native), العربية (native), Español (beginner)
side project

Lucent

In 2026 I started building Lucent, an iOS app meant to be a calmer alternative to screen-time trackers: a coach that named the pattern behind late-night scrolling instead of just showing a number and making you feel bad about it.

I got as far as a working Swift/iOS prototype: Supabase accounts, a 7-step onboarding flow, four working app tabs, before deciding to stop. Lucent is discontinued, not paused. I'm leaving it here because it's work I did, not because I'm still building it.

Stack: Swift/iOS, Supabase, OpenAI

Lucent home screen: today's focus ring, a 4-day streak, and Nudge/Blocker mode toggle

home

Lucent's AI coach, Sol: names the week's pattern ("The Gap Filler"), flags late-night scrolling, and takes open-ended questions

coach

Education and certifications

Education

MSc Business Analytics & Data Science

IE University, Madrid2027 cohort

BCom, Business Technology Management, Minor in Data Intelligence

Concordia University, MontréalSep 2022 - Jun 2026

Certifications

Python Data Associate

DataCampVerify

Data Analyst Associate

DataCampVerify

LLM Engineering: Master AI, LLMs & Agents

Udemy, Mar 2025Verify

Quantitative Research Job Simulation

JPMorgan Chase & Co. via Forage, Oct 2025Verify

Get in touch

Hiring for data analyst or data scientist roles, or want to swap notes on a project? My inbox is open.

contact.json
{
"status": "open to data analyst & data scientist roles",
"github": "github.com/Wiss-21"
}
$open --contact