Maxim Pilipenko

Numbers have to earn my trust.

I’m a Data Science & AI student looking for a junior data, BI or business analyst role. I check that a number makes sense before it gets anywhere near a model.

Based in Karmiel, open to roles in the North and Center, as a student position or full-time.

Income 117,000,000
Most likely 117,000, recorded 1,000× too big
Amount borrowed 562,491
Credit-to-income 0.005population median near 3
Outcome Defaulted
Exhibit A, straight from the Home Credit data. Earning 117 million and defaulting on half a million would be quite the plot twist.

Experience

Operations & Inventory Manager

Comedy Store, a convenience store, 2023–2026

Yes, it was really called Comedy Store. The expiry dates were no joke.

  • Tracked sales and stock in the Aviv POS system, from purchasing to back-office reports.
  • Used min/max reorder thresholds to decide what to buy and when.
  • Returned stock close to its expiry date to suppliers for full or partial refunds, instead of writing it off.

Math tutor

Perach project, 2026

Project

Credit default risk on 58.4 million rows

Python, SQL and PostgreSQL, scikit-learn, XGBoost, Jupyter

  • Loaded the seven related tables of the Home Credit Default Risk dataset, 58.4 million rows in total, into PostgreSQL and indexed the join keys.
  • Judged extreme values by whether the rest of the record backed them up, not by a cut-off rule. That’s how Exhibit A turned up, along with 55,374 applicants “employed” for about 1,000 years: a placeholder for no current job, and 99.96% of them were pensioners.
  • Dropped 28 redundant housing columns (three versions each of 14 measurements, correlated at 0.97–0.998) and ten near-constant flags.
  • On the application table, baseline models reached a PR-AUC of 0.217 with logistic regression and 0.234 with untuned XGBoost, against a no-skill baseline of 0.0807.
  • Turned down a 0.002 “improvement” because cross-validation put the noise at about 0.017.

Smaller than the noise means it didn’t happen.

Code on GitHub The cleaning notebook

Toolbox

Python
pandas, NumPy, scikit-learn, XGBoost, statsmodels, Matplotlib, Jupyter
SQL
PostgreSQL
R
dplyr, ggplot2, time-series models
Spreadsheets
Excel and Google Sheets, at an advanced level
Slides and BI
PowerPoint, plus the basics of Power BI
AI assistants
Claude, Claude Code, ChatGPT, Gemini, Perplexity
Also
GitHub, and the Aviv POS retail system

AI writes my boilerplate. The analytical calls stay mine.

Education

Western Galilee College

B.A. student, specialising in Data Science & AI, 2025–2027

Coursework includes machine learning in Python, econometrics in R, statistics and data analysis, corporate finance and investments, and accounting.

Magshimim and StarTech

Two years each, 2019–2023

C, C#, Python, networking and Wireshark at Magshimim. Java and Arduino at StarTech.

Writing code since 2019.

Languages

Hebrew and Russian, native. English, fluent.

Plus Python, R and SQL for talking to data.

Hiring a junior analyst?

Message me on LinkedIn, or look through my code on GitHub first.