Open to collaborations

Hello, I'm Azharul Islam. Call me

AZH

I build the software telco and fintech risk teams run on: billion-row CDR pipelines, fraud and revenue-assurance controls, AML/CFT monitoring, automated reporting, and the consoles people work in all day.

01 · Now

Controls by day, tools by night.

Day job

Revenue assurance & fraud management

I sit inside a payment provider's RAFM and internal-control team, so I build for the people who chase leakage, fraud and compliance gaps every morning.

How I build

Self-hosted, air-gap friendly

Tools that install from one offline file, stay read-only against production data and keep a record of everything they do.

Off hours

Desktop apps and phone tooling

Arduino and ESP32 boards, Android kernel builds, Windows utilities, and speech-to-text experiments in Bangla and English.

02 · Work

Things I've shipped.

Preview

Nov 2026 · Windows app

BusyMouse

Keeps a Windows PC awake by gently nudging the pointer, so the screen doesn't lock and chat apps keep you available. Lives in the tray, toggles with Ctrl+Shift+J, and steps aside while you use the mouse.

  • C#
  • WinUI 3
  • Installer
Try it in the browser →

Oct 2026 · Identity · Fintech

eKYC QC Console

A new front end for a wallet's eKYC quality-control console. Same back end, session and permissions as before, and pages not rebuilt yet open inside the new shell, so nobody loses a tab.

  • React
  • TypeScript
  • Vite
Preview

Jul 2026 · Reporting

MailFlow

Corporate auto-email reporting. Compose rich HTML emails with live SQL tables and charts, attach CSV, XLSX and PDF exports from the same queries, and send them on a schedule through your own SMTP relay.

  • Oracle
  • PostgreSQL
  • Trino
  • Teradata
  • Cron
Read the handbook →

Compliance · Knowledge

Regulatory Repository

Harvests regulatory documents (circulars, guidelines, directives) from the regulators' sources, classifies each one against the requirements it touches, and keeps a searchable library compliance can rely on.

  • Harvesting
  • Classification
  • Search

Risk · Fintech

RAFM Solution

Revenue-assurance and fraud-management controls for a payment provider: control monitoring, KPI boards, and a Postgres mirror of the data lake that keeps every change even when the source rewrites history.

  • PostgreSQL 16
  • Hive / Trino
  • Control atlas

Compliance

AML/CFT Solution

Anti-money-laundering and counter-terrorist-financing monitoring: screening and transaction rules, alerts that land in a review queue, and an audit trail compliance officers can stand behind.

  • Rules engine
  • Case review
  • Audit trail

Productivity

Enterprise Email Client

A corporate mail client for teams that live in their inbox: shared mailboxes, fast search and the workflow features a busy operations desk needs.

  • IMAP / SMTP
  • Shared mailboxes
  • Search

Also built

  • azhDroidManage an Android phone from a browser tab over WebUSB: screen mirror, shell, files, apps.
  • SimBox ClassifierOutbound dialer that records each answer and sorts it into human, IVR or SIM box with speech recognition.
  • pg_psp_rafmA Postgres mirror of a Hive schema that stays up, keeps history and makes joins fast.

03 · Data at telco scale

About three billion CDRs, every day.

~3B

call detail records processed daily

Automated CDR processing end to end: collection, decoding, validation, loading and reconciliation. It runs unattended and someone notices when a single file goes missing.

Decoding

ASN.1 decoder builder

Generates decoders from switch vendors' ASN.1 specs, so a new CDR format takes a spec file instead of a rewrite.

Pipelines

ETL / ELT

Loads from switches, mediation and the data lake into warehouses and Postgres, with idempotent reloads and row-level lineage.

Glue

Bash automation

The scripts that move files, rotate logs, watch disks and page a human before anyone else has noticed.

Illustration · how a call record travels

From the switch to the dashboard

  1. Switch
  2. ASN.1 decode
  3. ETL
  4. Fraud check
  5. Hive / Trino
  6. Power BI
≈ 34,722 records a second, if ~3B a day were spread evenly ≈ 0 at that rate since you opened this page raw decoded flagged, set aside

A simulation, not a live feed. No real or company data is shown: the counters are simple arithmetic from a rough daily volume, and the share of flagged records is made up.

04 · Stack

What I reach for.

Languages

  • Bash / Shell
  • TypeScript
  • Python
  • C++ (earlier)

Front & back end

  • React
  • Vite
  • Node.js
  • FastAPI
  • C# / .NET

Data

  • Oracle
  • EDB Postgres
  • PostgreSQL
  • MySQL
  • Teradata
  • Hive / Trino
  • ETL / ELT
  • ASN.1

BI & reporting

  • Power BI
  • Tableau
  • MailFlow

Automation

  • Selenium
  • Power Automate
  • Bash
  • cron

Design

  • Figma
  • Adobe XD
  • Photoshop
  • Illustrator

Hardware

  • Arduino
  • ESP32
  • Android kernel

Linux & ops

  • RHEL
  • Oracle Linux
  • Debian / Ubuntu
  • Docker
  • Git
  • Cloudflare

05 · Contact

Got an idea that needs a tool?