◆ Samith.me — Personal website

Samith Wijesinghe

iOS Engineer & ML Researcher — building intelligence at the edge.

I'm Samith Wijesinghe, a senior iOS engineer and machine-learning researcher based in Colombo, Sri Lanka. I build native iOS applications in Swift and Objective-C and explore the frontier of on-device machine learning — fine-tuning models, optimizing inference for Apple silicon, and running edge AI with Core ML and Apple's MLX framework. I build intelligence that runs locally, on the device, in your hands — not in the cloud.

What I do

• iOS engineering — native apps in Swift, Objective-C, SwiftUI and UIKit. I maintain a single codebase that ships 400+ white-label restaurant apps on the App Store at Incentivio.
• On-device ML & edge AI — running ML models locally with Core ML and MLX on Apple silicon, LoRA fine-tuning, and inference optimization for mobile hardware.
• ML research — deep learning and time-series forecasting with LSTM neural networks (published, cited 19×).
• Robotics — soft-robotics and mechatronics background, including a bio-inspired quadruped robot.

Skills

Languages: Swift, Objective-C, Python, C++, JavaScript. Frameworks: SwiftUI, UIKit, Core ML, MLX, ARKit, Combine. ML / AI: PyTorch, ONNX, MLX, LoRA, LSTM, on-device inference, edge AI. Tools: Xcode, Fastlane, GitHub Actions, MATLAB, Figma, Git.

Selected projects

Incentivio — iOS Developer (2022 – present)

Native iOS app for a restaurant ordering and guest-management platform; one codebase powering 400+ white-label apps on the App Store across the United States.

Soft Robotics Quadruped — Research Assistant (2018 – 2020)

Quadruped robot using soft robotic feet as actuators; designed the locomotion control system and MATLAB simulations for bio-inspired walking gaits.

Writing on MLX, edge AI & on-device ML

• Running LLMs on iPhone with Apple's MLX framework
• Edge AI on iOS: why on-device beats the cloud
• MLX vs Core ML: choosing an on-device inference stack
• All blog posts →

Research

Time Series Forecasting: Analysis of LSTM Neural Networks to Predict Exchange Rates of Currencies — Instrumentation, Vol. 7 Issue 4 (2020), cited 19 times. Google Scholar.

Connect

GitHub · LinkedIn · Medium · Twitter / X · [email protected] · Download CV (PDF)