Skip to content

Available for remote workDhaka, Bangladesh

Samiul Islambuilds the stackand the models.

Full Stack & ML Engineer with 1 yr 10 mo of experience. I build robust, scalable backend services, highly interactive user-centric frontends, and the machine learning models that make them smarter.

bun run profile.ts
// last updated Sep 2026
export const samiul = {
  role: "Full Stack & ML Engineer",
  experience: "1 yr 10 mo",
  frontend: ["Next.js", "React", "TypeScript"],
  backend: ["Node.js", "Go"],
  data: ["PostgreSQL","MySQL", "MongoDB"],
  infra: ["Docker", "Kubernetes"],
  ml: ["PyTorch", "scikit-learn", "Hugging Face"],
  available: true,
};
  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • Shadcn UI
  • Node.js
  • PostgreSQL
  • MySQL
  • MongoDB
  • Docker
  • Kubernetes
  • Python
  • PyTorch
  • scikit-learn
  • Hugging Face
  • Pandas & NumPy
  • FastAPI

01/About

Ships production software, and the models inside it.

I'm a full stack and machine learning engineer with 1 year and 10 months of professional experience, shipping production software from database schema to pixel. I care about clean, maintainable code: small functions, honest names, and tests that document intent. The goal is always a codebase the next engineer can read on day one.

On the backend I design and optimize REST and GraphQL APIs, model data in PostgreSQL and MongoDB, and ship services in Node.js and Go that are containerized with Docker. On the frontend I turn those APIs into fast, accessible interfaces with Next.js, React, and TypeScript.

The machine learning side is where the two halves meet. I train and fine-tune models in PyTorch and scikit-learn, evaluate them honestly, and serve them as versioned APIs that live in the same pipelines and dashboards as the rest of the product. A model that never reaches users is a notebook; I build the ones that ship.

Based in
Dhaka, Bangladesh (GMT+6)
Experience
1 yr 10 mo, since Jan 2025
Currently
Full Stack & ML Engineer at Company Name
Focus
Next.js products with ML services behind them
Working on
Demand forecasting model served from a Nodejs gateway

Two disciplines / one engineer

01

Software Engineering

Products people use every day.

  • Next.js and React frontends with TypeScript, tuned for Core Web Vitals and accessibility.
  • REST and GraphQL APIs in Node.js and Go with clear contracts and integration tests.
  • PostgreSQL and MongoDB data modelling, migrations, and query optimisation.
  • Docker images and Kubernetes deployments with zero-downtime rollouts.

02

Machine Learning

Models that reach production.

  • Supervised learning for forecasting, classification, and ranking with PyTorch and scikit-learn.
  • Fine-tuning transformer models with Hugging Face for text classification and retrieval.
  • Feature pipelines in Pandas and NumPy, tracked and reproducible with MLflow.
  • Serving models as FastAPI services with monitoring for drift and latency.

02/Stack

A small toolkit I know deeply.

Everything here has been used in production, not just in tutorials. Four layers, one system.

  • Frontend

    Interactive, accessible interfaces that feel instant.

    • Next.js
    • React
    • TypeScript
    • Tailwind CSS
    • Shadcn UI
  • Backend & Runtimes

    Typed, observable services with clear API contracts.

    • Node.js
  • Databases & DevOps

    Schema design, containers, and reliable deployments.

    • PostgreSQL
    • MySQL
    • MongoDB
    • Docker
    • Kubernetes
  • Machine Learning

    Training, evaluation, and serving models in production.

    • Python
    • PyTorch
    • scikit-learn
    • Hugging Face
    • Pandas & NumPy
    • FastAPI
frontend / backend / data / mllive

03/Projects

Products and models, measured by outcome.

  1. 01SaaS platform2026

    SeloraX Messages

    Built a multi-tenant SMS gateway for Bangladesh end to end: organization-scoped auth and API keys, a credit wallet whose append-only ledger debits atomically before dispatch and refunds rejected sends, self-service top-ups verified server-to-server through the EPS payment gateway, delivery-report polling, per-recipient link tracking, and an audited admin panel. Exposes a versioned public REST API with idempotent retries and rate limiting.

    • Next.js
    • TypeScript
    • Prisma
    • MariaDB
    • Better Auth
    • Vitest
  2. 02Real-time AI platform2026

    SeloraX Voice AI

    Built a multi-tenant voice agent platform that answers real phone calls with a speech model I trained myself. A Node gateway streams microphone or SIP call audio to the model over one persistent WebSocket per call, with barge-in, live transcripts, and self-hosted LiveKit WebRTC rooms. Each client gets its own agent with a prompt, HTTP and client-side tools, lifecycle webhooks, a knowledge base, S3 call recordings, and a poisha-precise wallet with per-customer markup. Every turn's speech and text is captured to keep improving the model.

    • Next.js
    • TypeScript
    • Node.js
    • Custom speech model
    • LiveKit
    • WebRTC
    • MongoDB
    • S3
    SeloraX Voice AI screenshot
  3. 03Machine learning2026

    Bangla Voice Model

    Fine-tuned OmniVoice, a 0.6B speech model on a Qwen3 backbone, for Bengali with LoRA so the Voice AI platform speaks in two consistent voices. Built the full data pipeline from raw call audio: normalized Bengali transcripts so digits match what is spoken, trimmed and loudness-normalized clips to 24 kHz, tokenized them with a neural audio codec, then trained on a free-tier T4 with checkpoints compared by ear against a zero-shot baseline. The chosen adapter is merged into the base model and published to Hugging Face.

    • PyTorch
    • LoRA
    • OmniVoice
    • Hugging Face
    • Bengali TTS
    • Colab

04/Experience

1 year 10 months of professional growth.

From a first production deploy to owning full-stack architecture, shipping ML models into the product, and coordinating delivery across a team.

  1. Jan 2026 — Present

    Full Stack & Machine Learning Engineer

    SeloraX

    Multi-tenant SaaS platform for Bangladesh's e-commerce market

    • Designed and implemented a transactional credit-based billing engine using a reserve/confirm/rollback workflow with Inngest, ensuring data consistency during concurrent processing through an event-driven architecture.
    • Trained and deployed a PyTorch demand forecasting model as a FastAPI service, with MLflow tracking and drift alerts.
    • Led API contract reviews between frontend, backend, and data engineers, cutting integration bugs found in QA by half.
    • Integrated the Meta Conversions API with backend services to improve purchase attribution accuracy and enable reliable third-party API integrations.
    • Optimized backend workflows and database interactions to improve system reliability and maintainability in a production environment.
  2. Jan 2025 — Dec 2026

    Full Stack Developer

    Naturo BD

    E-commerce platform migrating from Laravel to a modern JS stack

    • Engineered the frontend migration from Laravel to Next.js, modernizing the application with reusable React components and TypeScript.
    • Built responsive, production-ready user interfaces with a focus on performance optimization, maintainability, and scalability.
    • Designed, optimized, and debugged complex MySQL queries against denormalized database schemas to support high-volume e-commerce operations.
    • Collaborated with backend engineers to define API contracts and integrate frontend applications with RESTful backend services.
    • Contributed to a 300% increase in sales following the Laravel-to-Next.js migration through faster page loads and an improved user experience.

06/Guestbook

Leave a doodle on the wall.

Every developer who stops by can pin a little drawing and a note. Be nice; the wall is shared.

Sign the guestbook

Draw a little something, sign it, done. No links, no tracking.

draw something here ✎

07/Contact

Have a role or a hard problem? Let's talk.

I reply within a day. Email works best, or use the form.

Sent straight to my inbox. Nothing else is stored.

↑ ↓ to move↵ to goT toggles the themeX x-ray? all shortcutsdrag on blank paper to doodle/ opens this anywhere

Keyboard shortcuts

Ctrl/⌘K
command palette
/
command palette, too
T
paper ⇄ chalkboard
X
x-ray: see the components behind the page
?
this sheet
drag
doodle on blank paper
Ctrl/⌘Z
undo the last doodle
Esc
close any dialog
↑↑↓↓←→←→BA
…you know what this does

psst: open the browser console too.