|
I'm a Software Engineer who loves building things that actually matter — apps people use every day, systems that are fast and secure, and AI that runs on the device, not just in someone else's cloud. My work spans the full spectrum: from pixel-perfect Android & Flutter apps, to on-device LLM infrastructure that survives offline, to reinforcement-learning-driven edge cloud orchestration, to serverless AWS backends, to systems-level tools written in Rust that take memory safety seriously. I've shipped a 500K+ download Play Store app, built Memo — a caching and on-device fallback layer for LLM APIs — researched PPO-based proactive pod migration for real-time edge cloud workloads, built a Women's Health AI on Compose Multiplatform, created a blockchain identity system with Ktor, and written a high-encryption file manager in pure Rust. Outside of work, I write about software engineering on Medium and enjoy a good game of table tennis. |
| Android · Flutter · CMP | LiteRT · Gemma · Offline-First | PPO · K8s · CRIU | AWS · Lambda · DynamoDB | Rust · Ownership · Zero-cost |
| Platform | Stack | Proficiency |
|---|---|---|
| Android | Jetpack Compose · Kotlin · CMP | ████████████ Expert |
| Cross-Platform | Flutter · Dart | ████████████ Expert |
| Cross-Platform | React Native · TypeScript | ██████████░░ Advanced |
| Shared UI | Compose Multiplatform (CMP) | ██████████░░ Advanced |
A caching + execution layer for LLM APIs on Android — cuts cost, survives offline, adapts to hardware.
Kotlin·LiteRT-LM·Room·Coroutines/Flow·KMP-ready core
Memo is built as three independent layers:
graph TD
App["Your App\ncalls memo.resolve()"]
App --> Memo
subgraph Memo["Memo Facade"]
direction TB
SM["MemoStateManager\nnetwork + model state"]
MC["MemoCache\ncaching + TTL"]
OD["OnDeviceFallback\nlocal LiteRT engine"]
end
Memo --> Core
Memo --> Lib
Core["memo-core\npure Kotlin\nhashing · TTL · cost"]
Lib["memo-android\nRoom · LiteRT · Network"]
style App fill:#1a1a2e,stroke:#9F7AEA,stroke-width:2px,color:#fff
style Memo fill:#0d1117,stroke:#00BFFF,stroke-width:1px,color:#fff
style SM fill:#161b22,stroke:#00BFFF,stroke-width:1px,color:#ccc
style MC fill:#161b22,stroke:#00BFFF,stroke-width:1px,color:#ccc
style OD fill:#161b22,stroke:#00BFFF,stroke-width:1px,color:#ccc
style Core fill:#161b22,stroke:#9F7AEA,stroke-width:1px,color:#fff
style Lib fill:#161b22,stroke:#9F7AEA,stroke-width:1px,color:#fff
Kubernetes·PPO / Deep RL·CRIU·Amazon EKS·Prometheus + Grafana· Private repository
MSc thesis research building a Proximal Policy Optimization (PPO) deep reinforcement learning framework that performs proactive, hardware-aware pod migration in Kubernetes-managed edge cloud environments. The system continuously monitors CPU, memory, and temperature via Prometheus and migrates workloads before resource bottlenecks occur, using CRIU (Checkpoint/Restore In Userspace) for near-zero-downtime live migration.
- 67.8% reduction in latency
- 63.8% fewer SLA violations
- 90% reduction in migration downtime
- Tested on real Amazon EKS clusters under mixed real-time workloads
- Bridges AI-driven decision-making with cloud-native orchestration for latency-sensitive edge workloads
Kotlin·LiteRT-LM·Room·JitPack·Hardware-Aware
A drop-in caching and execution layer that sits between any Android app and any LLM API. Identical prompts return instantly from a SHA-256/TTL-based cache, real dollar savings are tracked automatically, and when the network drops, Memo seamlessly falls back to a real quantized Gemma model running on-device via LiteRT-LM — not a stub, a working offline conversation.
- SHA-256 + TTL caching — zero-cost, zero-latency on repeat prompts
- Live network observability via reactive Kotlin
Flow - True offline fallback — on-device Gemma inference, no cloud required
- Hardware-aware model tiers (Lite/Standard) resolved by device RAM & storage
memo-coreis pure Kotlin, zero Android deps, fully unit tested, KMP-ready- Bring-your-own provider — OpenAI, Gemini, Claude, Groq, or custom
CMP·AI/ML·Android + iOS·E2E Encrypted
A cross-platform AI-powered women's health companion built with Compose Multiplatform (CMP). Tracks menstrual cycles, predicts fertility windows, flags anomalies, and delivers personalized reproductive health insights — empathetically and privately.
- AI-driven cycle analysis & personalized health insights
- Hormone trend tracking & ovulation prediction
- End-to-end encrypted health data
- Shared business logic via CMP (Android + iOS)
Jetpack Compose·Ktor·Blockchain·AWS DynamoDB
A cutting-edge Android app that lets users create, own, and share digital identity profiles secured on the blockchain. Backend runs on Ktor hosted on AWS.
- Decentralized identity ownership via blockchain
- Beautiful animated profile cards with Jetpack Compose
- QR code & deep link profile sharing
- Ktor backend with JWT auth + AWS DynamoDB + WebSockets
Rust·AES-256-GCM·CLI + TUI·Argon2
A high-security file management tool written entirely in Rust — available as both a CLI and an interactive terminal UI (TUI) powered by ratatui. Rust's ownership model guarantees zero memory leaks by design.
- AES-256-GCM file encryption with Argon2 key derivation
- Interactive keyboard-driven TUI via
ratatui - Zero-copy file streaming — no runtime overhead
- Secure vault creation, locking, and metadata management
Android·Kotlin·Jetpack Compose·Google Play Store
The app that started it all — a wallpaper app with over 500,000 downloads on the Google Play Store. Optimized image loading, curated HD categories, and a slick UI.
| Project | Stack | Highlights |
|---|---|---|
| Project Tracker Backend | Lambda · DynamoDB · S3 | Serverless, auto-scaling, CI/CD via GitHub Actions |
| Funds Management System | DynamoDB · S3 · SNS · EC2 · CloudWatch · Lambda | Real-time alerts, automated reconciliation, full monitoring |
| CI/CD Pipelines | GitHub Actions · Jenkins · Docker | Microservices deployment automation |
| Certification | Status |
|---|---|
| AWS Certified Solutions Architect – Associate |


