Staff Engineer - Mobile, Desktop & KMP

Role details

About Sarvam Sarvam is building the bedrock of Sovereign AI for India. The company is developing India’s full-stack sovereign AI platform, building across research, models, infrastructure and applications with a singular focus on making AI genuinely work for India. Sarvam works with leading enterprises and public institutions and is backed by Lightspeed, Peak XV, and Khosla Ventures. Sarvam partners with India’s leading brands, including Tata Capital, SBI Life, CRED, IDFC, and LIC. About the Role Sarvam's consumer and enterprise products — Indus, Kivi, and the products we build next — live on the devices people actually use: Android phones, iPhones, Windows laptops, Linux desktops, and Macs. These are not thin web wrappers. They are native applications that run on-device inference, capture and stream audio, render streaming model output, and keep working when the network drops. Across India's hardware spectrum — from a low-end Android phone on a patchy 2G connection to a developer's Mac on fibre — the app has to feel fast and reliable. We are hiring a Staff Engineer to own the client application layer across mobile and desktop, with Kotlin Multiplatform (KMP) as the unifying bet. This is an IC role. You will set the architecture, the patterns, and the bar for how Sarvam ships native apps across five platforms without five separate codebases. You will work across the Sarvam product stack — Indus, Kivi, and new products as they come — and partner closely with backend, model serving, and product teams rather than against a finished spec. What You’ll Do - Own the end-to-end architecture of Sarvam's mobile and desktop applications, built on a Kotlin Multiplatform core that shares business logic, networking, storage, and model-client code across Android, iOS, Windows, Linux, and macOS, with Compose Multiplatform and Jetpack Compose as the primary UI layer. - Ship and evolve Indus and Kivi as native apps on all five platforms — chat and streaming UI, voice capture and playback, on-device and cloud model invocation, offline-first behaviour, and background processing. - Lead the KMP adoption strategy: decide what lives in common code versus what stays native, design the expect/actual boundaries, and build the shared modules (networking, persistence, audio pipeline, model client, telemetry) that every product team builds on. - Build for India's real device and network conditions: low-RAM Android devices, intermittent connectivity, code-switching across 22+ Indian languages, and audio pipelines that must handle VAD, compression, and streaming under poor networks. - Integrate Sarvam's model APIs — ASR, TTS, translation, LLM, vision — into the client, including streaming responses, partial result rendering, and graceful degradation when a model call fails or times out. - Drive performance and reliability across platforms: cold-start time, memory footprint on low-end devices, audio latency, battery discipline, and crash-free rate. - Set the engineering standard for the client team: CI/CD across all five platforms, automated testing on real devices, code review culture, and a documentation-first approach to architecture decisions. - Decide what UI and platform code stays shared versus native — Compose Multiplatform where it buys reach, but native UI frameworks (SwiftUI, Win32, GTK) and native libraries on any platform where a shared abstraction would cost performance, correctness, or platform feel. - Partner with backend and inference teams on API contracts and streaming protocols — you shape them, not just consume them — and with product and design to ship coherent, native experiences rather than platform-agnostic ports. - Mentor engineers across the client team on platform-native depth (Android, iOS, desktop) and on KMP discipline, raising the floor of what the team can ship independently. What We're Looking For - 9+ years building production mobile and/or desktop applications that real users depend on, with strong ownership of what you shipped. - Deep Kotlin expertise in production, including coroutines, flows, and modern Android architecture; - comfort with Swift on iOS and with at least one desktop platform (Windows, Linux, or macOS). - Hands-on experience with Kotlin Multiplatform (KMP) on a shipped or in-production product designing shared modules, managing expect/actual boundaries, and dealing with the real friction of sharing code across mobile and desktop. - Strong command of native audio and streaming pipelines: audio capture, playback, buffering, VAD, codec handling, and low-latency streaming to and from model endpoints. - Experience shipping apps that work in constrained environments — offline-first, low-bandwidth, low-RAM — not just on a flagship device over fibre. - Solid networking and API integration instincts: streaming (WebSocket, SSE), retry and backoff, caching, optimistic updates, and debugging across the client-server boundary. - Hands-on experience with Compose Multiplatform and Jetpack Compose in production, and the judgment to drop to native UI frameworks and native libraries (SwiftUI on iOS, Win32 or GTK on desktop, platform audio or media APIs) when a shared abstraction would cost performance, correctness, or platform feel. - Familiarity with CI/CD and automated testing across multiple platforms, including real-device testing rather than emulator-only. - The judgment of a senior IC: when to share code vs. go native, when to abstract vs. inline, when to optimize vs. ship, and how to bring the rest of the team up with you. Bonus Points - Experience with on-device model inference or on-device ML (TensorFlow Lite, Core ML, GGML, or equivalent) on mobile or desktop. - Background in speech or audio processing — codec work, sample-rate handling, echo cancellation, or VAD — in a production app. - Experience building voice-first or dictation products, especially multi-language or Indic language voice experiences. - Desktop app shipping experience via Kotlin/Native or JVM-based deskto