Ayush Singh

Software Engineer · Mumbai, India · Hybrid

Trading platform — web product with 10k+ daily active users.

• Built the real-time data layer of a single-screen scalping platform on the web trading product: low-latency WebSocket market streaming (RxJS) and browser-side persistence (LokiJS/IndexedDB), alongside SciChart charting with custom indicators.
• Led the Kotlin Multiplatform migration of the flagship Android app (15k+ DAUs), mentoring a junior iOS engineer — reached 90–95% shared code (shared business logic + Compose Multiplatform UI), now running in the production Android app with zero regressions, and shipped the company's first iOS app (TestFlight beta) in under 3 months.
• Migrated the company's Android-only KYC onboarding SDK to KMP: rebuilt 60+ screens across 20+ flows from XML to shared Compose Multiplatform, bridging platform-only KYC/e-sign SDKs (Setu, Hyperverge, Leegality) via expect/actual declarations, and re-architected to a canonical shared KMP architecture (shared HTTP clients, Hilt → Koin DI).
• Designed the agentic AI development workflow that powered the migration — Claude Code agents with custom skills for planning, TDD-based migration, code review, and validation loops — adopted by teammates and reinforced with structured peer reviews.
• Built and shipped a SEBI employee trade-declaration compliance platform solo in under a week using the AI-native workflow: automated parsing and validation of trade reports across 6+ brokers (Zerodha, Groww, Angel One, ICICI, Dhan), cutting turnaround from days to minutes; adopted company-wide.

Software Engineer · Mumbai, India · On-site

Sports-tech platform — race events, timing, and media at national scale.

• Conceived and built a face-recognition race timing system as the sole engineer — prototype to live-event production on a portable phone-and-tripod setup, validated at 95% accuracy head-to-head against RFID at a 2,000-runner event while cutting per-event timing costs from ₹200,000 to under $5 of compute.
• Owned Captured, a web platform for managing, discovering, and selling race photos, end-to-end as its only engineer: frontend, backend, and a serverless photo-processing pipeline handling 1M+ photos monthly for 150k+ peak-season visitors, plus an organizer platform serving 100+ organizers across 1,000+ events monthly.
• Co-built a VPC-internal communication platform consolidating notification logic from all 3 company products: 50k+ email/SMS/WhatsApp/push notifications daily with multi-provider auto-retry (AWS SES, Twilio, Meta WhatsApp API).
• Cut peak-load page times on India Running (10k+ DAUs) from 30–60s to ~10s by independently implementing read/write datasource separation and tuning hot queries with raw SQL.
• Designed and shipped serverless media systems: on-demand photo watermarking with live organizer preview, personalized per-runner race videos (AWS Lambda, S3, Remotion), and a concurrent-upload CLI that turned a 12–20 hr, 3-person manual process into a 4 hr single-machine job for 150,000 photos (400 GB).
• Integrated a full observability stack across 10+ services (SigNoz, Sentry, AppOptics) and led security hardening — VAPT testing plus backend-issued S3 signed URLs, eliminating credential exposure for 10k+ users.