AI/ML Data Engineer · Automation Expert · Python · JavaScript

I build automation that survives real work.

AI agents, scraping systems, backend APIs, and data workflows built for the failure cases that appear after the demo.

Bangladesh · UTC+6 · Working remotely worldwide

Upwork profile$30K+Total earningsTop RatedUpwork talent
Selected proof

Built in public. Used in real workflows.

All selected work
02
Open-source data utility

Google News Decoder

Python and Node.js tools that resolve encoded Google News links back to their original publisher URLs.

View
03
Structured data extraction

Facebook Page Scraper

A Python package for extracting structured public page information without a WebDriver or API key.

View
System thinking

From messy input to an owned system.

Most automation projects do not fail because someone cannot write code. They fail because the source changes, an API returns bad data, a session expires, or nobody knows why the job stopped.

01
Messy inputSites, files, forms, APIs
02
Automation logicBrowser, code, model, rules
03
Checks + retriesValidate, retry, escalate
04
Useful outputAPI, database, sheet, alert
05
MonitoringLogs, health, ownership
Programming languages

One core language. Others where they fit.

Python leads the work. Go supports lean backend services and tooling. JavaScript and TypeScript extend systems into integrations and product surfaces. Rust remains a deliberate learning track.

01 / PRODUCTION / DAILYPrimary language

Python

The main tool for automation logic, scraping systems, APIs, bots, data pipelines, and operational utilities.

AutomationScrapingFastAPIData workflowsBots
02 / BACKEND / TOOLINGSupporting language

Go

A supporting language for lean backend services, command-line tools, and network-aware systems with straightforward deployment.

Backend servicesCLI toolsNetworkingConcurrency
03 / SYSTEMS / PRODUCTSupporting languages

JavaScript + TypeScript

Used when workflows need Node.js services, browser-side behavior, integrations, or a product interface around the automation.

Node.jsExpressPuppeteerIntegrations
04 / LEARNING / NOT EXPERTLearning track

Rust

Learning for performance, memory safety, and dependable backend tooling—not presented as production expertise.

PerformanceMemory safetyBackend tooling
DevOps / infrastructure delivery
Your app. Your server. Production-ready.

From GitHub to production with Docker or a direct deployment behind Nginx—secured, observable, and running on infrastructure you control.

01 / SourceGitHub
02 / DeployDocker / Native
03 / RouteNginx
04 / RunLinux / Ubuntu
05 / OperateSSL · DNS · Logs · Monitoring
Capabilities by outcome

Tools follow the failure mode.

01

Make repetitive work ownable

AI agents, document flows, scheduled jobs, notifications, and human review where interpretation helps.

OpenAI · Claude · Gemini · Ollama · structured output · orchestration
02

Collect data from difficult sources

Browser automation and scraping systems designed around sessions, changing markup, unreliable responses, and protected sites.

Python · Playwright · SeleniumBase · Scrapy · curl_cffi · proxies
03

Connect the systems around it

Backend services, APIs, webhooks, queues, bots, and database workflows that move work to the next useful place.

FastAPI · Node.js · REST · PostgreSQL · Redis · Telegram · Discord
04

Host and operate the application

Take an application from GitHub to production behind Nginx—using Docker or a direct native runtime—with Linux hosting, SSL, logs, monitoring, backups, and production troubleshooting.

GitHub · Docker · native runtimes · systemd · Nginx · Linux · AWS · Hetzner · DigitalOcean · DNS · SSL
Professional experience

Production work, not isolated demos.

Full timeline

DMP

Automation Engineer

Current automation engineering role with DMP.

D & M Bumper Exchange

Data Scraping

Ongoing data scraping work for an automotive-parts business.

Legend Motors

Data Specialist

Data specialist role supporting an automotive business in Dubai.

Working method

Prove the difficult part early.

I use AI where interpretation helps. I use normal code where the result must be predictable.

  1. 01Understand the real workflow
  2. 02Identify the failure risks
  3. 03Prove the difficult part
  4. 04Build and validate
  5. 05Deploy with visibility
  6. 06Monitor and maintain
Latest writing

Notes outside the build log.

Visit the blog ↗
Start with the broken part

What workflow is your team tired of babysitting?

Send the target website, integration, file flow, or current failure. I’ll reply with the first risk I would test.

Discuss a workflow