Usmanjon Abduraimov
Men haqimda

Salom, men Usmonjonman.

Data Scientist & Analytics Engineer · Python · ML · Django · BI

Tartibsiz maʼlumotlar bazasini qarorlarga aylantiraman — keyin esa bu qarorlarni har kuni beradigan ilovani quraman.

  • Long-Term Resident of Japan (定住者 Teijūsha) — no visa sponsorship needed
  • Kaggle Notebooks Expert — peak global rank #1,919
Rezyumeni yuklab olish (PDF) Chop etiladigan rezyume Menga yozing
100% of daily & monthly regulatory reports automated Kapitalbank
2 analysts freed from routine manual work Kapitalbank
>50% less manual analysis & reporting time Shinhan Accounting
20,000+ auto-graded questions in a platform I built Powerty

Mening yoʻlim

Ishni Alokabankda kredit tahlilchisi sifatida, kredit hujjatlarini birma-bir oʻqishdan boshladim. Keyin Davlat soliq qoʻmitasida yuk xatlari tizimini qogʻozdan onlayn shaklga oʻtkazishda qatnashdim. Yaxshi loyihalangan maʼlumot oqimi tashkilotni qanday oʻzgartirishini oʻshanda birinchi marta koʻrdim: ishlov berish vaqti ikki baravar qisqardi, hech kim ortiqcha ishlamadi.

Koreya hukumati granti meni Koreyaga olib bordi: Keimyung universitetida bir yil koreys tili (TOPIK 6), soʻng KAISTda Techno-MBA — u yerda Python, maʼlumotlar bazalari va mashinali oʻqitish kundalik vositamga aylandi. Seulda Shinhan Accounting va Dasan DMC uchun tahlil tizimlarini qurdim, Kapitalbankda esa risk boʻlimining butun tahlil platformasini yolgʻiz oʻzim — Oracle soʻrovidan dashbordgacha — yaratdim.

Kechqurunlari Powerty’ni — shu saytni — qurdim: Oʻzbekistondagi oʻquvchilarim men orzu qilgan kursga ega boʻlsin, dedim. Bugun bu minglab darslar va savollar, Telegram bot, haqiqiy ballash tizimli mock imtihonlar va javoblarning toʻgʻriligini tekshiradigan koʻplab testlar.

Nega Yaponiya

Menda Yaponiyada Teijūsha (uzoq muddatli rezident) maqomi bor — viza jarayoni yoʻq, qaysi ishda ishlashimga cheklov ham yoʻq. U yerdagi kuchli muhandislar jamoasiga tizimni boshidan oxirigacha olib borish tajribamni olib kirmoqchiman va u bilan uzoq yillar oʻsmoqchiman. Yapon tilini oʻrganyapman.

自己紹介 — yapon tilida

はじめまして。ウスマンジョン・アブドゥライモフと申します。

データ分析・データ基盤エンジニアとして、銀行・税務・会計の分野で、データの抽出から分析、ダッシュボード、Djangoによる社内システムの構築まで一貫して担当してきました。

定住者の在留資格を持っており、就労制限はありません。日本で長く働きたいと考えています。日本語は現在勉強中です。

Email abduraimovusmanjon@gmail.com Menga yozishning eng tez yoʻli LinkedIn linkedin.com/in/usmanjon-abduraimov Ish tarixi va tavsiyalar GitHub github.com/usmanjohn Kod — shu platforma ham Powerty powerty.uz Men loyihalab qurgan taʼlim platformasi Tableau Public Tableau · usmanjon.abduraimov Dashbordlar va vizual tahlil Kaggle kaggle.com/usmanjon ML va statistika daftarlari Telegram @Albetta Toʻgʻridan-toʻgʻri yozing YouTube @powertyuz Men tayyorlagan animatsion darslar Instagram @powerty.uz Qisqa videolar va jumboqlar

Nimani quraman — boshidan oxirigacha

Bir kishi, butun jarayon — men topshirgan deyarli har bir tizimning shakli shu.

  1. Olish Oracle · PostgreSQL · MySQL
  2. Tozalash Pandas · Polars
  3. Tahlil va model statistics · scikit-learn
  4. Vizualizatsiya Plotly · Dash · Tableau · Power BI
  5. Ilovani ishga tushirish Django · auth · roles
  6. Avtomatlashtirish scheduled reports · docs

Kaggle’da data science

Tanlangan daftarlar: savol, qanday yechildi va natija.

ExpertKaggle Notebooks tier
#1,919peak global rank, Notebooks
Top 12%today: #6,851 of 59,993
490forks of my notebooks
Machine learning 444 views · 14 forks

California house prices — from linear regression to a tuned random forest

Savol
How well can district features predict the median house value?
Qanday
  • Found 7 feature pairs correlated above 0.85 and replaced them with engineered ratios (rooms per household, people per family).
  • Stratified the train/test split on ocean proximity, so the 5 island districts could not all fall on one side.
  • One-hot + standard-scaling pipeline; linear regression as the baseline, then a random forest tuned with 5-fold grid search.

Natija: R² 0.60 (linear) → 0.78 (tuned forest, validation) → 0.72 on the untouched test set; RMSE ≈ $61k.

  • scikit-learn
  • GridSearchCV
  • StratifiedShuffleSplit
  • ColumnTransformer
Daftarni ochish
Statistics 577 views · 47 forks

Is home advantage real? A hypothesis test on the Premier League

Savol
Do teams score more at home than away, at the 5% level?
Qanday
  • Stated H₀ / H₁ for a one-tailed, two-sample test on all 380 matches of 2021–22.
  • Derived the t-statistic and p-value by hand from the formula, then checked it against SciPy.

Natija: t = 2.22, p = 0.013 < 0.05 — reject H₀: home sides score significantly more (1.51 vs 1.31 goals).

  • SciPy
  • t-test
  • Pandas
Daftarni ochish
Data quality & geo 1,211 views · 83 forks · medal

215,000 Indonesian schools — cleaning, enriching and mapping

Savol
Where are the schools, and who runs them — the state or private owners?
Qanday
  • Scraped 2020 census population from Wikipedia and joined it by province, fixing the names that did not match.
  • Caught impossible coordinates (a longitude of 9,720) and kept only rows inside Indonesia’s real bounds — about 3,000 dropped.
  • Mapped school density against population; compared public and private schools by stage and province.

Natija: A medal-winning notebook, forked 83 times. Jakarta turned out to have nearly as many private schools as public ones — public schools dominate everywhere else.

  • Pandas
  • read_html
  • Seaborn
  • geo scatter
Daftarni ochish
Time series 741 views · 29 forks

South Korea’s demographics, 2000–2022

Savol
What is happening to births, deaths and marriages, region by region?
Qanday
  • Monthly series for 17 regions, resampled to years and compared across birth, death, marriage and divorce rates.

Natija: Deaths overtook births nationwide in 2020; Sejong is the only region still growing naturally.

  • Pandas resample
  • Matplotlib
Daftarni ochish
Analytics & Plotly 125 views · 12 forks

What really drives an air fare? Interactive price analysis

Savol
Why does the same route cost different amounts on different airlines?
Qanday
  • Held route, class and number of stops fixed to isolate one driver at a time, instead of comparing raw airline averages.

Natija: On Mumbai → Delhi the number of stops, not the departure time, was the main driver of the price gap.

  • Plotly
  • pivot tables
Daftarni ochish

Kaggle’dagi barcha 14 ta daftar

Ish tajribasi

Rezyume ingliz tilida — men ishlaydigan tilda.

  1. Risk Analytics & Data Platform Engineer

    Kapitalbank JSCB · Tashkent
    02/2025 – Present
    • Designed and built, as the sole engineer, a secure internal Django analytics platform used by management and the risk department.
    • Built the end-to-end pipeline: Oracle → preprocessing → validation → aggregation → analytics → dashboards → reports.
    • Automated 100% of daily and monthly regulatory and management reports, freeing two analysts from routine manual work.
    • Delivered portfolio analytics, NPL tracking, trend and growth modelling and KPI monitoring, behind role-based access.
    • Wrote the full technical and data-processing documentation. The system runs only on the bank’s internal network; a demo on synthetic data is available on request.

    Python · Django · Pandas · Polars · SQL · Oracle · PostgreSQL · Plotly

  2. Data & Cost Analytics Engineer

    Dasan DMC · Seoul
    10/2024 – 02/2025
    • Built a cost-modelling and profitability analytics system for automotive parts.
    • Automated data collection, transformation and scenario analysis.
    • Owned the data logic, models and analysis end to end.

    Python · Pandas · SQL

  3. Analytics Automation Engineer (Contract)

    Shinhan Accounting Corporation · Seoul
    02/2024 – 10/2024
    • Built a Django analytics system end to end: data ingestion, validation, processing, analytics and reporting.
    • Cut manual analysis and reporting time by more than 50%.
    • Built Tableau dashboards for client scoring, portfolio quality and risk indicators.
    • Owned the architecture, backend and data logic.

    Python · Django · Pandas · SQL · Tableau

  4. Risk Analyst

    State Tax Committee of Uzbekistan · Tashkent
    2020 – 2021
    • Led the digitalisation of the waybill system, from paper to online.
    • Designed the data workflow logic and validation structure.
    • Cut processing time by ~50% and made the process transparent.

  5. Credit Analytics Analyst

    Alokabank · Tashkent
    2019 – 2020
    • Prepared analytical summaries and risk indicators for 200+ loan cases.
    • Developed a structured evaluation logic for credit decisions.

Loyihalar

Powerty — data-driven learning platform

Founder & engineer · 2026 – Present
powerty.uz
  • Designed and built a bilingual (English/Uzbek) platform for GMAT, SAT, IELTS, TOPIK, Japanese and school maths: 1,300+ lessons, 20,000+ auto-graded questions, 800+ graded readings and timed mock exams.
  • Engineered an adaptive digital-SAT simulator (module routing, raw → 400–1600 scaled scoring) and a GMAT Focus simulator on one shared exam engine.
  • Built content pipelines guarded by automated “answer gates” — scripts that recompute every answer key by an independent route (brute force over fractions, state-space search) before import.
  • Runs in production on PostgreSQL + Railway: a cron-driven Telegram bot, an installable PWA, full i18n, TTS audio and headless-Chrome PDF generation.

Python · Django · PostgreSQL · AWS S3 · Git/GitHub · Railway · Bootstrap · AI-assisted development (Claude Code)

Kaggle — Notebooks Expert

14 public notebooks · 5 medals · 490 forks
kaggle.com/usmanjon
  • House-price regression (20,640 districts): treated multicollinearity with engineered ratio features, stratified split, one-hot + scaling pipeline; tuned a random forest with 5-fold grid search — R² 0.60 (linear) → 0.72 on held-out test.
  • Home-advantage hypothesis test (380 Premier League matches): one-tailed two-sample t-test derived by hand and checked against SciPy — t = 2.22, p = 0.013.
  • Indonesian schools (215,372 rows): joined census data scraped from Wikipedia, caught impossible coordinates and filtered ~3,000 to valid bounds before mapping.
  • South Korea demographics 2000–2022 (17 regions, monthly): time-series resampling showed deaths overtaking births nationwide in 2020.

scikit-learn · SciPy · Pandas · Plotly · Seaborn · Matplotlib

Powerty raqamlari bazadan jonli olinadi.

Koʻnikmalar

Data & analytics

  • Python (advanced)
  • SQL
  • Pandas
  • Polars
  • NumPy
  • scikit-learn
  • Statistics & A/B testing
  • ETL pipelines
  • KPI & portfolio analytics

Backend & ops

  • Django (auth, sessions, roles, ORM, i18n)
  • PostgreSQL
  • Oracle
  • MySQL
  • REST-style APIs
  • Git / GitHub
  • Railway
  • AWS S3
  • Cron & task automation

BI & visualisation

  • Tableau
  • Power BI
  • Plotly
  • Dash
  • Matplotlib
  • Report generators

Domains

  • Banking risk (NPL, portfolio, KPI)
  • Tax administration
  • Cost & profitability
  • Credit analysis
  • EdTech

Taʼlim

Techno-MBA

KAIST — Korea Advanced Institute of Science and Technology · South Korea
2022 – 2024
  • Korean Government Scholarship (GKS)
  • GMAT 710
  • Big Data Analysis, Database Management, Financial Data Analysis, Machine Learning, Business Analysis

Korean Language Program

Keimyung University · South Korea
2021 – 2022
  • Korean Government Scholarship (GKS)
  • TOPIK Level 6 — the highest level

Bachelor of Taxation

Tashkent Financial Institute · Uzbekistan
2015 – 2019
  • Thesis: tax strategies of the USA, Korea and Germany, applied to the Uzbek tax model

Tillar

  • Uzbek Native
  • English Fluent — professional
  • Korean Fluent — TOPIK 6
  • Russian Intermediate
  • Japanese Studying

Ishdan tashqari

  • Two-time chess champion of Syrdarya Province, Uzbekistan
  • Creator and narrator of animated educational videos (YouTube, Instagram)
  • Teaches Korean, English and mathematics

Yaponiyada xodim izlayapsizmi?

Menga yozing — Kapitalbank platformasining sintetik maʼlumotlardagi demosini soʻrov boʻyicha koʻrsataman.

Rezyumeni yuklab olish (PDF) abduraimovusmanjon@gmail.com