Geo
Native apps for iPhone, iPad and Mac for an open knowledge platform, with on-device AI and Web3 transactions.
role
Principal Mobile Engineer
platforms
iOS · iPadOS · macOS
worked with
CEO, Design and Backend
period
Sep 2023 – Aug 2026 · remote
// How it fits together
A simplified view of the app architecture
Drawn from my own description of the work. No proprietary code, screens or data.
Views
SwiftUI + UIKit · iPhone, iPad, Mac
↓
Resources
answer from local state, then revalidate
↓
Repository
one access point to the graph
↓
Sync actor
TTL policy · connectivity · stale-write protection
↓
Local store on LMDB
SwiftSyntax schema · query engine · background model actors
⇄ network
GraphQL and APIs
The sync actor reconciles remote updates with what the user already sees, without blocking the interface.
on-device AI
Apple Foundation Models + RAG
A lightweight retrieval pipeline over local data, so answers stay on the device.
web3
ERC-4337 transactions
Wallet authentication, bundler integration, gas sponsorship and ABI calldata encoding.
// Context
Instant on device, in sync with the network
Geo is an open, decentralised knowledge platform built on graph-backed data. The native apps had to respond immediately from local data, stay consistent with remote changes and support on-chain transactions.
// My contribution
Architecture, data and transactions
I owned technical direction and architecture across iOS, iPadOS and macOS, working directly with the CEO, Design and Backend from product decisions to TestFlight and production.
I designed Resource, Repository and Sync-actor layers so the app answers from local graph state and safely reconciles remote updates, TTL expiry and connectivity changes. When SwiftData no longer met performance needs, I replaced it with a SwiftData-style layer over LMDB, with a SwiftSyntax-generated schema, query engine and background model actors.
With no prior Web3 experience, I took ownership of mobile transactions: ERC-4337 account abstraction, wallet authentication, bundler integration, gas sponsorship and ABI calldata encoding.
// Highlights
AI inside the product and in the way we built it
On-device AI
Apple Foundation Models with a lightweight on-device RAG pipeline: retrieval and context design, output quality and product behaviour.
~90% less
manual effort on QA report triage, structured GitHub issue creation and first-pass code review, with AI-assisted workflows (internal estimate for these activities).
Local-first
Swift 6 concurrency with actors, background processing and stale-while-revalidate data flows.
ERC-4337
Account abstraction, wallet authentication and sponsored gas, verified byte by byte.
What are you working on?
Tell me about the product, the team or the technical question you need help with. Pick the time that suits you.