Kashif Mehmood

Senior Android & Kotlin Multiplatform Engineer
Islamabad, Pakistan (Remote, Global) +92 342 5332484 kashismails@gmail.com linkedin.com/in/kashif-mehmood-km github.com/Kashif-E x.com/kashif_mehmood_

Summary

Senior Android and Kotlin Multiplatform Engineer with 6+ years shipping mobile products to millions of users. Sole engineer on an enterprise on-device Vision AI SDK used by clients including WeWork; previously rebuilt a 10M+ download app in Compose Multiplatform at a $36M ARR AI company and scaled a Y Combinator (W22) startup's app solo to 5M+ users. Featured on the official Compose Multiplatform homepage; 2,600+ GitHub stars across open-source KMP libraries.

Technical Skills

Languages: Kotlin, Java, Swift, Python, Rust (working knowledge)

Mobile: Android SDK, Kotlin Multiplatform (KMP), Compose Multiplatform, Jetpack Compose, SwiftUI interop, Coroutines and Flow, MVVM, MVI, Clean Architecture, multi-module architecture, server-driven UI

On-Device AI / Computer Vision: TensorFlow Lite, LiteRT, custom CNN training, FP16 GPU inference, ML Kit, CameraX, Camera2, ARCore, OCR, barcode and document detection

Platform and Tooling: CI/CD (release automation), Gradle, ProGuard/R8, Firebase (Analytics, Crashlytics), Stripe SDK, REST and streaming APIs, Play Store and App Store delivery

Practices: SDK design, Kotlin DSLs, performance profiling, crash triage (99%+ crash-free), team training and onboarding

Experience

PackageXNov 2025 - Present
Senior Software Engineer, Vision SDK (returned)New York, USA (Remote)
  • Sole engineer owning the Vision SDK end to end, from model training to release: an AI-powered logistics SDK for Android, iOS, and React Native shipped to enterprise clients including WeWork.
  • Trained a custom CNN barcode detection model optimized for FP16 GPU inference; improved from ML Kit's 2-3/10 detections at 220-300ms to near-perfect detection at 40-60ms - a 5-7x latency improvement that makes real-time scanning viable on mid-range devices - using multi-stage, multi-resolution training without NMS.
  • Built a QuadNet-based document detector achieving clean edge detection in 40-60ms, a capability ML Kit does not offer, differentiating the SDK for enterprise buyers.
  • Designed a Kotlin DSL for on-device model lifecycle management (download, load, unload, infer) used across the full SDK.
  • Led migration from TensorFlow Lite to LiteRT following Google's deprecation of TFLite; enhanced CameraX with Camera2 interop for faster capture and finer pipeline control.
Vyro (Imagine Art and EverAsk, $36M ARR, 10M+ users)Nov 2024 - Oct 2025
Senior Software Engineer, Kotlin MultiplatformRemote
  • Rebuilt Imagine Art (10M+ downloads) from scratch in Compose Multiplatform in 2 months, moving iOS and Android onto one codebase so features ship to both platforms at once; achieved native iOS performance at a consistent 30-40fps with a clean multi-module architecture (app/feature/core/data/domain).
  • Launched EverAsk (formerly Chatly), a new AI agent and chat product, from zero to Play Store in one month; architected a streaming LLM SDK abstracting multiple providers behind a single domain interface.
  • Built a Compose-first streaming markdown renderer from scratch; achieved a 50% render performance improvement via recomposition scoping, derivedStateOf, and stable wrappers around mutable token state.
  • Built the revenue path: Stripe checkout in Compose covering subscriptions, one-time purchases, and trial paywalls, with full state restoration so no purchase is lost mid-flow.
  • Ported internal libraries to KMP and built the base architecture, sample app, and starter template used across all new KMP projects; trained 3 engineers in KMP from a View-system background.
Markaz Technologies (Y Combinator W22)Jul 2022 - Oct 2024
Mobile Engineering Lead (Android, iOS, KMP)Islamabad, Pakistan
  • Scaled a social commerce app to 5M+ users as the sole mobile engineer for 1.5 years, targeting low-end devices (1-2GB RAM) with unpredictable OEM behavior; maintained a 99% crash-free rate and a top-5 category ranking throughout.
  • Introduced Jetpack Compose and led the full XML-to-Compose migration, then introduced KMP and delivered the Compose Multiplatform iOS launch ahead of industry adoption.
  • Built a separate KMP + SwiftUI supplier-side app in 2022 - one of the earliest production KMP + SwiftUI implementations.
  • Pitched, designed, and built a server-driven UI framework end-to-end (backend and mobile), letting the business team launch seasonal campaigns (Eid, sales) on the home screen same-day, without app releases.
  • Hired and trained a team of 5 engineers in KMP and Compose Multiplatform; built onboarding docs, internal sessions, and a sandbox codebase for hands-on learning.
Markaz TechnologiesDec 2021 - Jul 2022
Founding EngineerIslamabad, Pakistan
  • Built and shipped the entire Markaz MVP in 14 days - categories, search, orders, shipments, payments - a key factor in Y Combinator W22 acceptance; reached 200K downloads and 10K MAU within 3 months.
  • Launched the Markaz Supplier app in 7 days using Kotlin Multiplatform Mobile - among the earliest production KMM apps in Pakistan; cut app size from 22MB to 3.8MB via ProGuard and R8, widening reach to suppliers on low-storage devices.
DockYard, Inc.Nov 2022 - Apr 2023
Software EngineerBoston, USA (Remote, Part-Time Contract)
  • Founded the LiveView Native Android SDK, enabling Elixir developers to build native Android UIs with Jetpack Compose, bridging the Phoenix web ecosystem with native Android.
  • Built a Rust library for HTML/DOM tree traversal, improving parsing performance by 60%.
Telenor Microfinance Bank (Easypaisa)Sep 2021 - Dec 2021
Software EngineerIslamabad, Pakistan
  • Delivered core billing, KYC, and bill payment flows on Pakistan's largest digital payments platform (10M+ active users); shipped 3 major on-schedule releases with test coverage on critical payment and onboarding flows.
PackageXFeb 2021 - Sep 2021
Android DeveloperIslamabad, Pakistan
  • Built Sight, a real-time people counting and headcount analytics system using live camera feeds, shipped to WeWork offices; built ARCore experiences for Staples.
  • Improved the mailroom app's crash-free rate from 92% to 99%+ for enterprise customers and reduced memory usage by ~80% via proper bitmap lifecycle management during high-frequency image capture.
  • Integrated Firebase ML Kit OCR with CameraX, boosting image recognition accuracy by 40%.
9D TechnologiesFeb 2019 - Jun 2020
Android DeveloperPakistan

Open Source & Community

Education

University of the PeopleIn progress
Bachelor's degree, Computer ScienceOnline, USA