Performance under pressure. Agents under control.

ZZAZZJan 2026 – Present

Senior Mobile Engineer — Bootstrapping React Native for iOS & Android from zero

Bootstrapped the company’s React Native app for iOS and Android from zero.

Built the Exchange feed as native Kotlin + Swift list views on Nitro HybridViews — each row redraws its Canvas graphs and live metrics on its own cadence.

Wrote the team’s agent tooling: a worktree-isolated task runner whose reviewer subagent grades each run on diff, logs and screenshots, then emits the next-iteration prompt as a self-improving loop. Plus an MCP server for driving a live build.

Owned context engineering for the repo: an AGENTS.md rule set and reusable agent skills that keep agent-written code on convention.

Built an in-app AI assistant on a ReAct loop whose dynamic tool registry lets the model navigate screens and fill forms, streaming from three providers.

Added on-device text-to-speech with ExecuTorch for offline article read-aloud. Implemented CI/CD with GitLab pipelines, caching, dual release targets, and automated submission.

ZeptoNov 2024 – Dec 2025

Senior Mobile Engineer — Consumer quick-commerce · 40M+ MAU

Led performance optimisation across the consumer app: 35% faster screen loads, 25% less memory, higher frame rates on low-end devices.

Built live order tracking with maps for single and batched orders — from partial coverage to 100% of orders trackable in real time.

Shipped gamification surfaces (scratch cards, floating elements) and monetisation modules — 20% lift in promotion-driven conversions.

Built an in-house notification SDK: +18% campaign CTR and removed the external vendor entirely.

Built an internal mock-management tool that decoupled feature work from backend readiness. Integrated image-classification pipelines for automated product validation.

Stanverse TechnologiesJan 2022 – Oct 2024

Mobile Engineer — Creator platform · 3.5M+ monthly users · team of 6

Built three real-time React Native games reaching 50K+ daily gameplays, plus an audio-room platform for 50K+ daily users on Agora + WebSocket.

Raised crash-free sessions from 94% to 99.3% with Crashlytics.

Reworked login flows for a 40% lift in success rate; integrated UPI and card payments handling 5K+ daily transactions.

Flagship — Stan × Upstox, IPL 2022: single-handedly built and deployed the responsive IPL 2022 campaign site for Upstox, scaling to 500K+ daily active users.

EWar GamesAug 2021 – Dec 2021

Software Engineer — Internal tooling

Built an admin panel with CRUD, filters, search, pagination and auth.

Established a role-based permission system across internal tooling.

ZZAZZ Exchange FeedZZAZZ

Problem. Live market rows need Canvas graphs and metrics that redraw on independent cadences without janking the list.

Approach. Native Kotlin + Swift list views on Nitro HybridViews; each row owns its own Canvas redraw loop.

Outcome. Buttery feed scroll with per-row live graphics.

React Native · Nitro Modules · Kotlin · Swift · Canvas

Self-Improving Agent Task RunnerZZAZZ

Problem. Agent runs needed isolation, graded feedback, and a loop that actually improves the next prompt.

Approach. Worktree-isolated runner + reviewer subagent scoring diff, logs, and screenshots, emitting the next-iteration prompt.

Outcome. Closed-loop agent shipping with reviewable diffs.

Claude Code · MCP · Subagents · Git worktrees

In-App ReAct AssistantZZAZZ

Problem. Users needed an assistant that could navigate screens and fill forms — not just chat.

Approach. ReAct loop with a dynamic tool registry; streams from three providers with on-device TTS via ExecuTorch.

Outcome. Tool-calling assistant that drives the UI.

ReAct · Tool calling · ExecuTorch · Multi-provider

Live Order TrackingZepto

Problem. Only a subset of orders were trackable in real time across single and batched deliveries.

Approach. Maps-based tracking surfaces for single and batched orders, wired end-to-end into the consumer app.

Outcome. 100% of orders trackable in real time.

React Native · Maps · Realtime

Performance ProgrammeZepto

Problem. Screen loads and memory pressure were hurting conversion on low-end devices at 40M+ MAU.

Approach. Cross-cutting profiling, render path cuts, and memory discipline across the consumer app.

Outcome. 35% faster loads · 25% less memory · higher FPS.

React Native · Profiling · Reanimated

Stan × Upstox IPL 2022Stanverse

Problem. Upstox needed a responsive IPL campaign site that could absorb cricket-scale traffic.

Approach. Single-handedly designed, built, and deployed the campaign site under a tight Mar–May window.

Outcome. Scaled to 500K+ daily active users.

React · Responsive web · Campaign