Orbit 02 / Selected work
●Software engineering, machine learning & product systems
I like software that makes complicated things feel calm: clear inputs, honest outputs, and an architecture that holds up when the idea gets real.
Working principle / Build the smallest useful system first. Then make every layer explainable.
A retrieval workflow for exploring customer feedback through both semantic similarity and relationships between the things people mention.
Worked on the retrieval approach, knowledge-graph context, query routing, and evidence-led response design.
A fraud-detection exploration focused on the signals that make a payment feel unusual: velocity, amount, location context, and behavioral patterns.
Built the feature pipeline and evaluation workflow for transaction-risk classification.
SwiftUI product work that combines on-device capabilities, persistent data, and interfaces designed to feel at home on iPhone.
Prototyped the client architecture, interaction patterns, and device-first information flows.
Full-stack web experiences where motion and visual craft are part of the product, supported by a practical React and Next.js foundation.
Designed and built responsive interfaces, interaction systems, and front-end architecture.
Practice notes
01The machine-learning work is about decision support and careful evaluation—not pretending a model is magic.
02I care about performance because waiting makes an otherwise good interaction feel careless.
03A good prototype exposes the hard questions early enough to change direction.
From the beginning
The full story, from birth to today: the experiences, people, learning, and milestones that shaped my journey.
Full timeline coming later
Read my story