Abbasgulu Allahverdili — Data Analyst

About me

Data Analyst based in Baku, Azerbaijan. I work across the full data lifecycle — collecting, cleaning, exploring and modelling data with SQL, Python, Excel and Power BI — turning it into dashboards that answer real business questions.

How I Work

1Data Collection

Databases with SQL, the web with Python scrapers — automated to run itself.
2Data Cleaning

Missing values, duplicates and messy formats fixed in Python and Excel — so the numbers hold up.
3Exploratory Data Analysis

Patterns, outliers and relationships surfaced in Power BI dashboards you can read at a glance.
4Modelling & Algorithms

Statistical methods and algorithms that explain what drives the numbers — and predict what's next.

Resume

Education

Yildiz Technical University

BSc. Mathematical Engineering (2018 — 2022)

A degree built on statistics, linear algebra, numerical methods and programming. It's where I learned to turn a vague problem into a model, and a model into an answer someone can act on.

Experience

Business Analyst (Intern)

Bir Ecosystem — Baku, Azerbaijan (08/2026 — Present)

Collecting and analysing data to surface customer needs, business trends and KPIs, and supporting the reports and presentations that go to senior management. Also involved in requirements gathering, process mapping and workflow analysis for new projects.
Teaching Assistant (Intern)

Israel Azerbaijan Training Center (02/2023 — 05/2023)

Supported the instructor on a data analytics course: walked students through Python, SQL and statistics in the lab, reviewed their analyses, and broke down the concepts they got stuck on.
Data Analyst (Intern)

International Training & Project Center (09/2021 — 03/2022)

Cleaned and analysed large datasets using statistical techniques, identified trends to generate actionable insights, and designed reports and visualisations for stakeholders.

Certifications & Awards

Data Analyst Professional Certificate

IBM (07/2026)
Show credential
Data Analytics

Handex (07/2026)
Show credential
Data Science with Python

Data SoCool (05/2024)
Show credential
Programming Essentials in Python (PCAP)

Cisco Networking Academy · OpenEDG Python Institute (03/2023)
Show credential
Investigation of Brain Tumor Segmentation Using U-Net Architecture

TÜBİTAK · 2209-A Research Projects Programme (11/2022 — 11/2024)
Show credential

Skills

Python pandas, requests, BeautifulSoup
SQL joins, aggregation, window functions
Power BI DAX, data modelling, dashboards
Excel pivot tables, lookups, reporting
Web Scraping automated collection pipelines
Git version control

Languages

Download full CV (PDF)

Portfolio

Projects I've built — data collection, analysis and dashboards.

HR Workforce Analytics

Oracle SQL & Python — End to End
- Oracle 18c
- SQL
- Python
- Data Engineering
A synthetic 10,000-employee organisation built in Oracle, then analysed end to end. The heavy computation — window functions, hierarchical CONNECT BY queries, percentiles — runs in the database; Python handles interpretation. Every query lives in sql/ as a reviewable file, not a string in a notebook.

What the analysis showed
- Promotions are driven by opportunity, not just performance — one department promotes 14.5% of staff simply because it has 6.6 juniors per senior seat.
- A fixed 24-month observation window flips which hiring cohorts look "under-promoted" — the framing changes the finding.
- Rewrote one hierarchy query from a correlated subquery to a pre-aggregated CTE: 85.6s → 0.06s, ~1,400× faster, same result.
GitHub repo Analysis notebook
Crimes in Boston

EDA & Severity Prediction
- Python
- pandas
- scikit-learn
- Folium
An end-to-end look at ~319,000 Boston crime records (2015–2018): what happens, when, and where — then a Random Forest testing whether time and place alone can predict how serious an incident is. The hard part wasn't the model; it was Latin-1 encoding, partial years and invalid coordinates.

What the data showed
- Most "crime" isn't violent — top calls are vehicle response, medical aid and investigations. Serious offenses are only ~19% of the data.
- Crime has a rhythm: it peaks on summer evenings (4–6pm) and stays active through weekend nights.
- Violence concentrates — shootings are 0.32% of incidents, but 74% hit just 3 districts, overwhelmingly late at night.
- Using only time and place (no offense type, to avoid leakage), the model reached ROC-AUC 0.60 — a modest but real signal.
Dataset Kaggle notebook
Credit Card Churn

Interactive Power BI Dashboard
- Power BI
- Power Query
- DAX
- Slicers
A dashboard exploring why credit card customers leave — and where retention effort pays off. Power Query for cleaning, DAX for the measures, and slicers that filter the whole report by gender, card, and income. 10,127 customers analysed.

What the data showed
- Overall churn is 16.07% — but customers inactive for 3 months churn the most, an early-warning signal worth acting on.
- Low-income customers (under $40K) churn hardest, and churn drops sharply as people hold more products — cross-selling is a retention lever.
- The 46–55 age group carries the highest churn count.
Dataset Project file
Oxu.az

News Dataset — 52,946 articles
- Python
- Web scraping
- EDA
- Kaggle
Local-language datasets are scarce, so I built one. A Python pipeline walks the archives of 13 categories and collects the title, date, author and reader reactions for every article published since 2013 — 52,946 articles in total.

What the analysis showed
- Sports is the most written-about category (5,812 articles), but Society is the most read — editorial focus and audience interest don't line up.
- In culture, tourism, ICT and show business the median article gets zero likes. Sports and politics run about three likes per dislike. People read the quiet categories; they just don't press the button.
- Short titles (1–4 words) hold a slight engagement edge — smaller than the folklore suggests.
Dataset on Kaggle EDA notebook
Global Superstore

Interactive Excel Dashboard
- Excel
- Power Query
- Power Pivot
- DAX
A fully interactive dashboard on the Global Superstore dataset: Power Query for cleaning and transformation, Power Pivot and DAX for the model and measures, PivotCharts and slicers for the front end.

What the data showed
- Central leads on sales at $2.8M; North Asia has the best margin at 20%.
- Technology drives profit even though Office Supplies moves more units — volume and profit are not the same story.
- Southeast Asia sits at a 2% margin. That's the number worth a meeting.
Project file Dataset

Contact

Every dataset has a story. Tell me yours.

How I Work

1Data Collection

2Data Cleaning

3Exploratory Data Analysis

4Modelling & Algorithms

Education

Yildiz Technical University

Experience

Business Analyst (Intern)

Teaching Assistant (Intern)

Data Analyst (Intern)

Certifications & Awards

Data Analyst Professional Certificate

Data Analytics

Data Science with Python

Programming Essentials in Python (PCAP)

Investigation of Brain Tumor Segmentation Using U-Net Architecture

Skills

Languages

HR Workforce Analytics

Crimes in Boston

Credit Card Churn

Oxu.az

Global Superstore

Contact Form