Reduan Nur
Labid .
Building mobile apps, web platforms, and ML systems that real people use.
CS graduate from BRAC University. I've shipped a multi-platform streaming app with 50+ organic users, built full-stack web platforms for real clients, and published ML research on GANs, computer vision, and NLP. I care about performance, reliability, and things that actually work.
Flagship Release • v3.0.0PotFlix Multi-Platform
A production-grade, hardware-accelerated streaming suite built for high-throughput networks (BDIX & Nagordola CDN). Available natively on Android, Windows, and Linux, backed by an automated GitHub Actions CI/CD release pipeline.
Home & Dynamic Catalog
Curated trending content, continuing-watching row, and live BDIX peering connectivity indicators.
Production-Grade Kotlin & Jetpack Compose
Designed from the ground up utilizing Clean Architecture & MVVM to decouple network I/O, local caching, and stateful UI:
- Jetpack Compose UI: Declarative UI powered by Compose BOM, Material 3 theming, Navigation Compose, and custom gesture controllers.
- Dependency Injection & Persistence: Dagger-Hilt with KSP for compile-time safe injection, paired with Room Database SQLite for instant offline metadata resume.
- Hardware Streaming Engine: Integrated LibVLC for Android with custom multi-audio decoding, subtitle synchronization (.srt/.vtt), and dynamic buffer policies.
- Baseline Profiles & R8: Generated custom Baseline Profiles to minimize JIT compilation, delivering near-instant warm startup latency and 60fps scrolling without ANRs.
Selected Works & Engineering Projects.
Machine learning research publications, production mobile applications, full-stack web platforms, and automated CI/CD pipelines.
Dual-Stage Manifold Purification and Adversarial Representation Learning: A WGAN-GP Critic Pipeline for Highly Imbalanced Intrusion Detection
Automated network intrusion detection struggles with severe class imbalance, where benign flows can outnumber rare attacks by up to 153,910:1. This research proposes a dual-stage, manifold-aware pipeline utilizing Fuzzy C-Means clustering for benign purification and tiered generative augmentation (CTGAN/TVAE) across 16.1 million flows. A WGAN-GP Critic network is adversarially trained and repurposed as a 128-dimensional latent feature encoder.
CineCompose
CineCompose is a framing assistant designed to provide actionable camera-composition guidance to cinematographers in real-time. To bypass the need for subjective and expensive manual labeling, the system utilizes a novel perturbation-based dataset pipeline, synthesizing supervised training data through controlled directional crops and minimal-scale signed rotations of professional stills.
Drug-Target Affinity (GNN & ESM-2)
State-of-the-art Drug-Target Affinity (DTA) prediction architecture integrating Geometric Deep Learning with Large Protein Language Models. Fuses an ESM-2 (650M) transformer with Graph Attention Networks (GAT) via Bi-Directional Cross-Attention.
Emotion Detection in Banglish
Novel Dual-Stream Contextual Transformer architecture for code-mixed Banglish text. Processes normalized Banglish via XLM-ROBERTa and transliterated Bengali via BanglaBERT, augmented by a 6D lexical probability vector and Focal Loss.
PotFlix Multi-Platform
High-performance media streaming client for BDIX & Nagordola CDN networks. Features LibVLC hardware acceleration, multi-audio/subtitle track management, Room offline caching, and automated GitHub Actions multi-arch CI/CD pipeline.
BD Public Pulse
Empirical 2D Political Compass & real-time public opinion analytics platform calibrated for Bangladeshi civil society. Features a 24-question ideological assessment, Euclidean distance leader matching, live Firestore presence & opinion aggregation, and downloadable 9:16 social opinion cards.
ShakalakaTv Live Streaming
Proof-of-concept web aggregator exploring live television channels and sports broadcast feeds. Demonstrates real-time match fixture ingestion via TheSportsDB REST API, multi-source HLS video stream decoding, channel categorization, and Cloudflare edge caching.
PetHood Mobile
Modern pet adoption mobile application built using MVVM clean architecture. Features high-performance 'Reels-style' vertical scrolling, instant intent dialer integration, Room local persistence, and real-time breed filtering.
Multi-Platform CI/CD Release Pipeline
Automated end-to-end delivery pipelines configured with GitHub Actions. Automates multi-architecture Android APK builds (ARM64, ARMv7, x86_64), automated desktop packaging (Windows setup/portable & Linux), SHA-256 integrity verification, and Firebase Firestore synchronization.
Skills & Engineering Disciplines.
A comprehensive overview of production tooling, architecture patterns, automation pipelines, and core computer science competencies.
Native Mobile (Android)
- Kotlin (Production-grade)
- Jetpack Compose & Material 3
- Clean Architecture & MVVM
- Coroutines & StateFlow
- Dagger Hilt & KSP Injection
- Room Database & Local Caching
- LibVLC Hardware Streaming
- Baseline Profiles & R8 Optimization
APIs & Full-Stack Web
- REST APIs (TMDB, IntroDB, TheSportsDB)
- API Rate Limiting, Caching & Pipelines
- TypeScript & Modern JavaScript
- React, Next.js & Astro Ecosystems
- Node.js & Express API Development
- PostgreSQL & Supabase (RLS/RPCs)
- State Management & Data Fetching
- WebSocket & Client-Server Sync
Cloud, Hosting & CI/CD
- Cloudflare (DNS, Edge CDN, Caching)
- Cloud Hosting (Vercel, Netlify, GitHub Pages)
- Firebase Cloud (Firestore, Auth, Storage)
- GitHub Actions Workflow Automation
- Multi-Arch Build Matrix (ARM64/v7/x86)
- Automated Release & Checksum Generation
- Semantic Versioning & Changelogs
- Linux Environments & Shell Automation
Applied AI, ML & Research
- PyTorch & Deep Learning Models
- Graph Neural Networks (GNN / GAT)
- Protein Language Models (ESM-2)
- Transformer Models (BERT / RoBERTa)
- Generative Adversarial Nets (WGAN-GP)
- Edge Model Inference (983K flows/sec)
- Fuzzy C-Means Clustering & Data Synthesis
- Computer Vision & Multi-Head CNNs
App Health, Testing & Quality
- Startup Latency & Render Profiling
- Zero-ANR & Memory Leak Elimination
- Unit Testing (JUnit, Mockito)
- Instrumented UI Tests (Espresso)
- Network Resilience & Offline-First
- Crash-Free User Metric Monitoring
- API Contract Design & Review
- Code Review & Modularization
AI-Native Engineering & Tools
- Agentic IDEs (Cursor, Antigravity, Claude Code)
- Context Engineering & Prompt Optimization
- AI-Assisted Automated Test Generation
- RFC & Technical Spec Drafting
- Android Studio & Profiler Suites
- Postman & OpenAPI Specifications
- VS Code & JetBrains IDE Ecosystem
- Agile Sprint Planning & Delivery
Open Source Profile.
Live stats from GitHub, synced with recent open-source activity.
Top Languages
Featured Repositories
This is streaming app that is directly connected to the bdix ftp servers and fetches streams instantly
Bilingual 2D Political Compass & real-time public opinion analytics platform for Bangladesh
NLP & multimodal classification models for phonetic Banglish social text
Graph Neural Networks & ESM-2 protein language representations for DTA prediction
github.com/ReduanNurLabid
View contribution heatmap, commit history and pinned research
About My Background.
I build things I'm genuinely proud of. Here's a bit about who I am and how I work.
I'm Reduan Nur Labid, a software engineer and CS graduate from BRAC University. I've spent the last few years building things across mobile, web, and machine learning — a streaming client that runs on three platforms with 50+ organic users, full-stack web platforms for real clients, and ML research on GANs, computer vision, and NLP.
I take performance seriously. Whether it's profiling an Android app to hit zero ANR, optimizing a model to run at 1.4 million inferences per second, or setting up CI/CD that deploys reliably on every push — I care about how things actually run, not just whether they work on my machine.
I'm also genuinely interested in Bangla NLP and code-mixed text, which I think is an underexplored space. If you're working on something at the intersection of language, ML, and real-world impact, I'd love to talk.
Academic Foundation
B.Sc. in Computer Science
Graduated with core focus in Software Engineering, Mobile Systems, and Applied AI.
Higher Secondary Certificate (HSC)
Science discipline. Built strong foundational mathematical, logical, and algorithmic problem-solving skills.
Core Engineering Principles
- Performance matters end-to-end: I profile Android apps for zero ANR, optimize ML models for edge inference, and measure before I claim anything is fast.
- Automate the boring parts: CI/CD, release tagging, multi-platform builds — if it can run on a push, it should.
- Ship things people use: Research is great. But the best feedback comes from real users with real problems — that's what I optimize for.
Open to Opportunities.
Looking for full-time roles in Mobile Engineering, Full-Stack Web, or AI/ML. Open to positions in Dhaka or remote. If something sounds like a fit, I'd love to hear from you.
Looking for a dedicated Software Engineer?
Whether you need a mobile engineer, a full-stack developer, or someone who can take an ML project from research to production, I'm ready to contribute from day one.