Dhaka, Bangladesh

Ahmed Afridee

Flutter Developer | AI Engineer

I build offline-first mobile systems and on-device AI pipelines — 5+ years shipping production Flutter apps, from architecture to the App Store.

scroll to explore
Portrait of Ahmed AfrideeOpen to Flutter & AI roles

01 — about

About

I’m a Flutter developer and AI engineer with 5+ years of production experience across Android and iOS. Today I build offline-first, high-volume systems that keep 4,500+ field agents working through zero-connectivity zones; before that I spent five years as the sole developer of a live employee management app used across client organisations. More recently my work has moved into applied AI: on-device pipelines built with LangChain, LangGraph, flutter_gemma, and RAG, carried from architecture through App Store deployment. I wrote about that work in Towards AI (2026), on running LLMs on-device in Flutter.

B.Sc. in Computer Science and Engineering

North South University · 01/2018 – 08/2022

5+

Years shipping production Flutter apps

4,500+

Field agents served by offline-first systems

99%

Reduction in sync errors on reconnection events

50,000+

Daily writes handled by the offline schema

02 — experience

Work Experience

Five years of production Flutter — offline-first architecture, sync engines, and apps maintained solo from first commit to App Store.

Manush Tech

Flutter Developer · Full-time

Singapore

04/2025 – Present

Eliminated order entry failures in zero-connectivity zones for 4,500+ field agents by building an offline-first capture layer with Drift — orders now complete successfully regardless of network state

Reduced sync errors by 99% and eliminated data loss across 50,000+ daily reconnection events by designing a bidirectional sync engine with deterministic conflict-resolution logic, co-owning API contracts with backend engineers

Architected a Drift-backed local schema supporting 15+ entity types and 50,000+ daily writes, enabling complete offline parity with the server-side data model — zero feature degradation without connectivity

Elements Group

Flutter Developer · Full-time

Sydney, NSW · Remote

02/2020 – 04/2025

Sole developer on a live employee management app, used by 257 employees across client organisations

Eliminated fraudulent attendance entries entirely — achieving 0 incidents post-launch and removing the need for manual supervisor sign-off — by building a GPS-verified clock-in system using Geolocator

Developed a dynamic checklist engine allowing managers to configure and assign day-specific task forms to field employees

Delivered real-time push notifications (FCM) for shift reminders, task alerts, and broadcasts — achieving 98%+ delivery reliability across iOS and Android

Architected the app with GetX for state management and Dio for REST API communication, maintaining the codebase solo for 5 years across 10+ releases

Configured Codemagic CI/CD pipeline for automated builds and TestFlight/App Store distribution, keeping manual deployment overhead at zero

Wrote unit and widget tests covering core business logic including geo-clock-in validation and checklist state management

03 — projects

AI Engineering Projects

Agents, retrieval pipelines, and on-device inference — built end to end and shipped.

project 01

Karo

Building an AI Agent That Knows Your Company

Built an AI agent that ingests internal knowledge (API docs, business rules, procedures), enabling team members to query live data in plain English — no SQL, Postman, or engineering tickets needed

Added an eval harness for regression testing retrieval accuracy and agent behavior across prompt/schema changes; fully containerized with Docker Compose

Implemented a hybrid RAG + API-calling architecture using LangGraph/LangChain, with pgvector for semantic retrieval and a dynamic HTTP tool supporting per-request auth injection

Persisted multi-turn conversation history in PostgreSQL via LangGraph's PostgresSaver; built chat UI in Chainlit with text and voice input (server-side transcription via faster-whisper)

PythonLangGraphLangChainOpenAI GPT-4opgvectorPostgreSQLChainlitDocker
Smart Notes architecture: on-device RAG with HNSW hybrid retrieval and Gemma 4 E4B IT

project 02

Smart Notes

Fully private, offline AI inference on-device

Achieved fully private, offline AI inference on-device — zero cloud calls, zero API keys — by integrating Gemma 4 E4B IT (~4.3 GB) and EmbeddingGemma via flutter_gemma for generation and 768-dimensional text embeddings

Implemented hybrid retrieval combining dense cosine search (HNSW via ObjectBox) and BM25 keyword scoring, fused with weighted-sum normalization (0.7 × dense + 0.3 × BM25) and MMR re-ranking for result diversity

Designed a semantic graph view using pairwise cosine similarity across note mean vectors, surfacing topical connections without tags or folders

Managed a fixed 2,048-token context window with dynamic token-aware prompt trimming across system instruction, retrieved chunks, and query

FlutterDartflutter_gemmaGemma 4 E4B ITEmbeddingGemmaObjectBoxHNSWBM25

05 — skills

Skills

Mobile foundations, AI and backend systems, and the tooling that ships them.

Mobile & Flutter

Flutter
Android
iOS
Dart
flutter_gemma
GetX
Provider
Riverpod
BLoC
Dio
Drift
SQLite
Geolocator
Push Notifications
Unit Testing
Widget Testing
Responsive & Adaptive UI
Firebase

AI & Backend

Python
LangChain
LangGraph
OpenAI GPT-4o
Gemini API
RAG
pgvector
PostgreSQL
Vector Search
Embeddings
ObjectBox
BM25
on-device ML
Multimodal
Chainlit
FastAPI
Docker
REST APIs
faster-whisper
prompt engineering
agent architecture
hybrid retrieval pipeline

Tools & Workflow

Git
GitHub
Codemagic CI/CD
Agile
Scrum

06 — contact

Let's build something that works offline too.

Open to Flutter and AI engineering work — reach out through any channel below.