Pinwheel Watch Suite
Full suite of mission-critical wearable apps — Launcher, Contacts, Messenger — supporting tens of thousands of active users on constrained hardware. Built with Jetpack Compose and MVI.
Senior Engineer @ Pinwheel · Mobile · Backend · ML · Product
I started in mobile — shipping scalable Android and WearOS apps with offline-first functionality and tight performance budgets — and have since grown across the stack: backend (TypeScript / NestJS), applied machine learning (CNN-based computer vision and speech-to-text systems), AI-assisted development workflows with Claude, and product decisions. I also ship my own apps on the Play Store. Generalist by output, specialist where it counts.
I'm a senior engineer based in Dhaka, currently working at Pinwheel (Austin, USA) where I build wearable experiences for tens of thousands of users — and increasingly contribute beyond mobile into backend, AI tooling, and product.
On the mobile side, my work lives at the edge of what hardware can do — squeezing battery life out of wearables, designing motion algorithms from raw accelerometer data, and making apps work offline in remote areas with no signal. On the backend, I work in TypeScript with NestJS supporting the Pinwheel Home product.
I'm also driving AI integration into our development lifecycle — using Claude for peer code review and other tasks — and partnering closely with the product team on feature design, requirements gathering, and ticket scoping.
Before Pinwheel, I led mobile development at Monstarlab, shipped Denmark's #1 podcast platform, and contributed to open-source Kotlin Multiplatform tooling. I spoke at KotlinConf'23 and won MonstarHacks 2024.
The ultimate kids habit builder & chore app — making everyday routines fun through rewards, photo verification, and room-based task tracking.
The only local firewall that can't spy on you. Take total control of mobile data, extend battery life, and protect privacy — all on-device, zero permissions.
An end-to-end machine learning application designed to empower deaf and mute individuals in education. Combines a custom-trained convolutional neural network for sign-language recognition with real-time speech-to-text — turning natural communication into a two-way bridge.
Curated, labeled, and augmented a 3,750-image dataset for sign language gestures — handling class balance, lighting variance, and background diversity for robust generalization.
Fine-tuned VGG16 on top of ImageNet pre-trained weights, freezing early conv blocks and retraining the classifier head for sign recognition with high accuracy on a small dataset.
OpenCV pipeline for hand region detection, frame normalization, and live inference — feeding processed frames into the CNN for low-latency on-device gesture classification.
Two-way communication: Google Speech API converts spoken language to text, while sign-language gestures are recognized and translated back to text — enabling fluid conversation.
Beyond research projects, I'm actively integrating Claude into our engineering lifecycle — automating peer code review and accelerating engineering productivity.
Spoke at KotlinConf '23 on production mobile engineering — bridging applied ML, mobile platforms, and developer experience for a technical audience.
Full suite of mission-critical wearable apps — Launcher, Contacts, Messenger — supporting tens of thousands of active users on constrained hardware. Built with Jetpack Compose and MVI.
Denmark's top podcast app for Berlingske Media. Re-architected playback with ExoPlayer, boosting retention by 28% and reducing crashes by 16%.
Cross-platform attendance and meal tracker built with Kotlin Multiplatform, reducing company food waste by 30%. Ships on Android, iOS, and WearOS.
Backend services supporting the Pinwheel Home product — built with TypeScript and NestJS. Cross-functional work alongside mobile, designing APIs and shipping features end-to-end.
Driving AI integration into our engineering workflow at Pinwheel — using Claude for automated peer code review and other repetitive engineering tasks to ship faster with higher quality.
Won the company-wide hackathon, partnered with Amazon Web Services (AWS).
Presented at KotlinConf'23 hosted at Brac University, Dhaka, Bangladesh.
North South University, Dhaka, Bangladesh · Apr 2016 – Dec 2020
I'm currently open to interesting opportunities — especially around mobile, wearables, or anything where performance matters. Drop me a line.