Open to internships and entry-level data or BI roles

Brishav Rajbahak / Data analyst in progress

I want the work to look good. I need the numbers to hold up.

An undergraduate student in Nepal, learning how to make sense of real datasets and the decisions behind them.

PythonCleaning and analysisSQLQueries and data structurePower BIDashboards and reportingCloudPages, Functions and delivery
Start with the proof
01 / Work

The projects, with the proof left in.

Published results show their definitions. Unfinished work says exactly where it stands.

Project mapFour projects. One relationship map.

Loan Default Analysis / Financial Inclusion Gap Analysis / Loan Default Prediction / Portfolio Platform

The mandala is the main interaction. The terminal is here only if you want the command-line version.

Project 01Published

Loan Default Analysis

A Lending Club analysis where the denominator mattered as much as the chart.

Business impact

  • Separates completed outcomes from active loans before measuring default.
  • Shows where grade, term and purpose carry different risk levels.

What I delivered

  • A reproducible Python and SQL cleaning workflow.
  • A two-page Power BI report covering the overview and risk segments.
19.98%Final-outcome default rate6.04%Grade A49.67%Grade G
Project 02Published

Financial Inclusion Gap Analysis

A completed comparison of Nepal's financial access gaps across account ownership and digital finance.

Business impact

  • Makes Nepal's account-ownership and digital-access gaps comparable with South Asia, lower-middle-income and world benchmarks.
  • Surfaces the population groups that still lag in the published comparison, including women, poorer groups and older adults in 2024 digital access.

What I delivered

  • A cleaned, analysis-ready Global Findex dataset and documented Python and SQL workflow.
  • Three Power BI report pages: Executive Overview, Nepal Account Gaps and 2024 Digital Access.
8,577Source observations18Analysis-ready indicators3Dashboard report pages
Project 03Published

Loan Default Prediction

An explainable loan-risk predictor that uses only information available when the loan is issued.

Business impact

  • Shows how origination-time loan details can be used to estimate historical bad-outcome risk without using post-loan information.
  • Makes the model's limits explicit: it is an educational, retrospective analysis and not a lending or credit-approval system.

What I delivered

  • A leakage-audited, time-based model comparison and 2018 holdout evaluation.
  • A live Streamlit predictor with dependent grade and sub-grade controls, responsible-use guidance and an estimated bad-outcome probability.
0.69122018 test ROC-AUC66.45%2018 test recall1,348,099Completed-loan rows
Project 04Live system

Portfolio Platform

This portfolio, built as a static-first frontend with protected edge functions.

Business impact

  • Gives reviewers one place to inspect projects, methods and working interfaces.

What I delivered

  • A static frontend, protected contact route and interactive dataset endpoints.
Published grade comparisonOpen in a new page

The dashboard is ready when you want to load it.

02 / Process

The denominator stays visible.

One Lending Club dataset changes shape as it moves from raw rows to the published 19.98% result.

Ingest

Raw records

2,260,701 accepted-loan rows enter the workspace.

1 / 6
01

Ingest

2,260,701 accepted-loan rows enter the workspace.

02

Cleanse

Types are repaired and 33 footer rows are removed.

03

Explore

Only loans with a finished outcome enter the rate.

04

Model

Grade, term and purpose are compared before adding complexity.

05

Visualize

The result becomes a report that can be inspected.

06

Impact

The published KPI keeps its numerator and denominator visible.

2,260,701raw records / 2007–20182,260,668cleaned loan rows1,348,099final-outcome loans269,360bad outcomes1,078,739good outcomes19.98%269,360 ÷ 1,348,099
Why 912,569 loans were excluded

912,569 loans were still current, late or in a grace period, so they did not yet have a final outcome.

Including unresolved loans would distort a rate that claims to compare final repayment with final default.

03 / About

This part is not polished on purpose.

It's me Brishav Rajbahak. Currently an undergraduate student learning new things and tools which i feel fascinating . i relate myself to the tech-enthuiast and wanna build cool things which not only exists but resonates well to the make an earth better place

Brishav's real workspace, photo 1
Loan-default dashboard work.
Brishav's real workspace, photo 2
A real working space, not a generated scene.
Brishav's learning routeA rising path connecting education, applied projects, business intelligence storytelling and the portfolio launch.
  1. FoundationEducation

    Computer science, statistics and the habit of checking the source.

  2. PracticeApplied Projects

    Python and SQL work built from real datasets rather than polished mockups.

  3. CommunicationBI Storytelling

    Dashboards shaped around questions, definitions and the person reading them.

  4. NowDataverse Launch

    Published projects and a frontend that keeps changing as the work improves.

being better than yesterday . just iterating the level even if its a word or an entire dictionary

always wanting best and supreme level of things i build . wanting to make things look preety as preety as the system works. progress is always > perfection , perfection has no limit and cant be achived

04 / Contact

If the work interests you, write to me.

The form is protected by Turnstile and rate limiting. You can also use one of the verified links below.

Kathmandu, NepalNepal Standard Time / UTC+05:45

The protected form loads as you approach it.