About
I'm Mulham Sarmini, a software engineer focused on full-stack development, machine learning, and shipping polished applications end to end.
I enjoy the full lifecycle of building products, from architecture and clean implementation to deployment and iteration.
Designing and building end-to-end web applications with React, Next.js, and modern backends.
Applying ML to real problems, from predictive models to on-device computer vision.
Taking ideas from prototype to deployed, maintainable apps on cloud platforms.
Skills
Showcase
A selection of applications I've designed, built, and deployed.
A hybrid fitness app that delivers live exercise-form feedback using on-device pose detection (TensorFlow.js + MoveNet). Built with Capacitor and a native Android wrapper, with an AI coaching chat, Supabase auth and cloud sync, and an offline fallback.
A procedural floor-plan and furniture arranging studio. A pure TypeScript SAT geometry engine powers real-time collision detection, door/window wall snapping, and simulated-annealing auto-arrange, with undo/redo, JSON/PNG/DXF export, a react-three-fiber 3D walkthrough, and an AI assistant for layout critique. Ships as a web app plus an Expo Android companion.
A smart ingredient-expiry tracker and recipe finder. Local-first food database with FoodKeeper-style shelf-life data, barcode scanning via Open Food Facts, and AI-generated recipes from Google Gemini routed through a Cloudflare Worker proxy to keep the API key off-device.
A full-stack subscription tracker that surfaces recurring spend and helps cancel unwanted services. Next.js web and Expo mobile apps with Supabase auth and per-user data, assisted cancellation that deep-links to real merchant cancellation pages, and a Playwright automation agent driven by a FastAPI + Celery backend.
An audio-to-notes synthesizer that turns recordings into structured summaries. A FastAPI pipeline chunks audio for parallel Whisper transcription and LLM summarization, with transcript search, chat-with-audio with timestamped citations, flashcards, share links, and a UI localized to 40 languages.
A personal health tracking application for monitoring your wellness journey, with mood tracking, symptom logging, medication reminders, and health analytics.
A machine learning system that predicts loan approval decisions with high accuracy, built using ensemble methods and deployed as a scalable REST API with a web interface.
Contact
I'm open to new roles, freelance projects, and collaborations. The fastest way to reach me is by email.
Send me an email