Analytics Engineer · Data Analyst

Less manual reporting, more decisions.

I'm Tasnim, an analytics engineer and data analyst with four years across fintech, travel, B2B and civic data. I turn messy sources into trusted numbers and automate the reporting, so teams spend their time deciding, not compiling. Currently at Gozayaan.

how_i_work select a stage
Runs near real-time · batch · daily, on GitHub Actions cron, n8n and scheduled queries Checked for duplicates · nulls · data types · row counts · reconciliation · failed-job email alerts
How I work: from a business question through ingestion, modelling, metrics and delivery to a decision. The loading method and schedule depend on the task.

What I do

Pipelines & integrations

I pull data from Google Cloud APIs, custom REST APIs, app databases, CSV files, sheets and forms, then land it cleanly in PostgreSQL or BigQuery.

Python · FastAPI · REST APIs · Google Cloud

Warehouse modelling

I write star-schema marts, production SQL and data-quality rules, so every team counts customers, transactions and revenue the same way.

SQL · BigQuery · PostgreSQL · Snowflake

Automation

Any report or data handoff that repeats becomes a scheduled job. Python does the work and cron or n8n runs it, so nobody has to remember.

GitHub Actions · n8n · Apps Script

Analytics & BI

Cohorts, retention, CAC and LTV, funnels, pricing and revenue, delivered as dashboards people use on their own.

Metabase · Power BI · Looker · Mixpanel

Recurring work, automated

If a report or data handoff happens on a schedule, it should run itself. I write the job in Python, schedule it on GitHub Actions or n8n, and deliver the output wherever people already look. I've used this pattern to retire hand-built daily reports and recurring data extracts.

Python · GitHub Actions · n8n · SQL · REST APIs

Results

A few outcomes from the work above.

Reporting speed

5 → 1 day

Month-end reporting turnaround after I redesigned the process.

Gozayaan

Public impact

~5,700 affidavits

Candidate affidavits from four national elections, scraped, digitised and published in Transparency International Bangladesh's public "Know Your Candidate" dashboard for voters.

Transparency International Bangladesh

Breadth

15+ dashboards

Dashboards and reports across 6 business functions, built on a warehouse I designed from the ground up.

Riseup Labs

Time saved

8–10 hrs / week

Manual data collection replaced with automated pipelines and n8n workflows into PostgreSQL.

Riseup Labs

Decision speed

30% faster

Shorter decision cycles once teams had live dashboards instead of waiting on manual reports.

Riseup Labs

Marketing funnel

+5–8% conversion

Funnel and channel analysis that found where prospects dropped off, and flagged a paid campaign that wasn't paying back, which was then stopped.

Riseup Labs

Four years, five teams

Pick a bar to read what I did there.

Selected work

Employer work is confidential, so these are personal projects. Each one covers the problem, what I did and what I found, with charts redrawn from the real output.

Python · pandas

Product sales, 12 months

Problem
A year of raw electronics orders from 10 US cities. When and where does revenue come from, when should ads run, and what sells together?
Approach
Cleaned the order data in pandas (blank and malformed rows, types) down to 186,305 orders, derived month, city and hour columns, then grouped orders and paired products bought in the same order.
Outcome
December is the best month at $4.61M and San Francisco the biggest city. Orders peak around 12 PM and 7 PM, so ads should run just before. A phone plus its charging cable is the strongest bundle.

Open the notebook ↗

SQL · MySQL

COVID-19 deaths & vaccinations

Problem
How hard did COVID-19 hit each country relative to its cases and population, and how fast did vaccination roll out?
Approach
Loaded Our World in Data CSVs into MySQL, joined the deaths and vaccination tables, and used CTEs, window functions and views.
Outcome
Fatality and infection rates per country, a rolling vaccination count for Bangladesh, and the top 20 countries by confirmed cases.
WITH cte AS (
  SELECT location,
    MAX(CAST(total_deaths AS UNSIGNED)) AS deaths,
    MAX(CAST(total_cases  AS UNSIGNED)) AS cases
  FROM coviddeath
  WHERE continent IS NOT NULL
    AND location = 'bangladesh'
  GROUP BY location)
SELECT location,
  ROUND(deaths / cases * 100, 2) AS fatality_rate
FROM cte;
Repository ↗
Power BI sales page with KPI cards, sales trend lines and category breakdown

Power BI · DAX · data model

Sales & customer analysis

Problem
Sales and customer data that was hard to read together across segments, countries, categories and time.
Approach
Cleaned in Power Query, modelled as a star schema, and wrote DAX for month-over-month change, margin and a rank-based customer segmentation. Two pages with year and category slicers.
Outcome
Loyal customers bring in 44.5% of sales. The USA and Germany are the top markets, Drinks and Cheese the top categories, and 2021 peaked before a slight dip in 2022.
Repository ↗
Dark Power BI dashboard showing revenue, cost, profit and margin by quarter and category

Power BI

Three-year sales dashboard

Problem
Three years of sales in separate sheets, with no single view of revenue, cost, margin or salesperson performance.
Approach
Appended the yearly sheets in Power Query, built the data model, and wrote DAX for cumulative sales, margin and profit against target.
Outcome
$25.66M revenue at a 32.5% margin, between 31.7% and 33.3% every quarter. Kids & Toys leads on profit at 43%, and the top seller by revenue was the only top seller to hit the profit target.
Repository ↗

More, including data engineering exercises and small tools, on GitHub ↗

Education & certificates

The credentials that shaped how I work, from degree to today. Most of them earned alongside the job.

off_the_clock hover to trace

Still learning

The rest happens after hours

Building small tools, learning AI and data engineering, and finding new trails.

Open to remote opportunities

Need reporting that runs itself?

I'm available for remote data engineer and analytics engineer roles, and I take on freelance projects in data pipelines, data modelling and dashboards.

Have a role or a data problem? Let's talk. Email is the quickest way to reach me.