I'm Minhazur Rahman — a data scientist, ML engineer & automation specialist. Everything I build is typed, tested, monitored, and live on the cloud — never a slide deck about what could ship.
3+ yrs data & infrastructure · MSc Data Science, University of Greenwich (London, UK) · now Chattogram, BD → relocation-ready
Two disciplines, one bar: production ML and human-in-the-loop automation. Every project is typed with mypy, tested with pytest, linted, CI-gated, containerised — and every metric comes from a real run, never a fabricated number.
Demand forecasting, fraud detection, and experimentation — honest baselines first, then models that have to beat them to earn a deploy.
Models served as real APIs on real infrastructure — IaC, keyless CI/CD, Prometheus metrics, and drift monitoring included.
Human-in-the-loop automations that do real work: document intake (OCR → validated ledger), support-ticket triage, RPA reconciliation, workflow orchestration, and job-fit scoring — free deterministic cores, with optional LLMs that degrade gracefully.
Orchestrated pipelines and real-time scoring — asset-based DAGs, tested warehouse models, Kafka consumers measured end-to-end in CI.
Three of these are live right now — score a transaction, forecast demand, or scan a careers page yourself. Every number comes from a real, reproducible run.
// live demos scale to zero ($0/month infra) — the first request after idle takes ~30s to wake the container; the demo pages let you watch it happen
Card-fraud detection on heavily imbalanced data: leakage-safe features shared between training and serving, threshold tuned on a held-out set, served with FastAPI + Prometheus, deployed via Terraform.
End-to-end demand-forecasting MLOps pipeline: typed, tested, CI-gated, containerised, with Prometheus metrics and Evidently drift monitoring — deployed keylessly through Workload Identity Federation.
Automation flagship: paste any company's careers page and get every open role scored 0–100 for your fit with a plain-English "why." 100% free, no API key — deterministic HTTP fetch, an optional local JS-render toggle, and multi-company compare.
Real-time transaction scoring over Kafka/Redpanda: alerting, dead-letter queue, Prometheus metrics, Kubernetes manifests with a kind smoke test — and CI that runs a real broker end-to-end.
An orchestrated ML platform: Dagster assets feed a tested dbt star schema on DuckDB, a LightGBM challenger races the champion, and a weekly automated retrain promotes it only if it clears the gate.
Document-intake automation: PDF invoices → OCR → Pydantic + arithmetic validation → DuckDB ledger, with a human-in-the-loop exception queue that catches anything malformed before it's booked. Free core loop.
After completing my MSc in London, I returned to Bangladesh and built a portfolio of production-grade systems solo to an industry bar: two live ML APIs on GCP Cloud Run (Terraform, keyless CI/CD), real-time Kafka scoring proven on Kubernetes, an orchestrated Dagster + dbt platform, and a suite of human-in-the-loop automation tools — and published a first-author, DOI-indexed preprint.
Pass with Merit (Sep 2024 – Jan 2026). Dissertation (70%): real-time retail demand forecasting on privacy-preserving synthetic data — a 37% MAE reduction, published as a DOI-indexed preprint on Zenodo.
Automated departmental reporting and operational workflows with Python and SQL at one of Bangladesh's largest listed steel manufacturers — internal tooling, data integrity, and systems administration.
Quality assurance for computer-vision training datasets — annotation review and data integrity at scale for 3D floor-plan reconstruction models.
Jan 2017 – Jan 2021. Final-year project: rare-event detection in surveillance video with 3D-ResNet18 and a custom CNN (PyTorch) on the UCF Crime dataset.
I'm actively looking for Data Scientist, ML Engineer & Automation roles — remote or on-site, anywhere in the world. Relocation-ready and visa-sponsorship friendly (EU HSM / Blue Card eligible). If my work looks like what your team ships, let's talk — I reply within a day.
$ echo "minhazurrahman.ds@gmail.com" | mail --subject "let's talk"