Frontend
- React
- TypeScript
- JavaScript (ES2017+)
- HTML5
- CSS3
- Marko
- Redux
- Vite
- Webpack
- Rollup
- SSR & hydration
Frontend / Full Stack Software Engineer · Greater Seattle Area
Tushar Tilwani — Frontend / Full Stack Software Engineer
I build fast, reliable, and observable web experiences at scale. Over 13+ years across eBay, Oracle Cloud, and Amazon, I've focused on React and TypeScript frontend architecture, frontend observability, web performance, and scalable frontend platforms that serve tens of millions of users a day.
- 13+years in big-tech frontend
- 2M+concurrent users supported
- 10M+daily commerce page views
- 100+teams on tooling I built
About
Senior / staff-level frontend engineer with deep big-tech experience
I'm a frontend-focused full stack software engineer with 13+ years building large-scale consumer and enterprise web applications. I currently work on eBay's Buyer Experiences team, where I operate at staff-level scope — driving frontend architecture, web performance, and cross-team technical standards for commerce pages that serve 10M+ views a day.
My core strength is the intersection of frontend architecture and production reliability: designing component systems and rendering paths in React, Marko, and TypeScript, then making them observable, fast, and dependable in production. I built a lightweight OpenTelemetry-based frontend observability framework adopted by 100+ engineering teams, architected a real-time chat platform supporting 2M+ concurrent users at sub-200 ms latency, and have a long track record of Core Web Vitals and web performance wins on high-traffic pages.
Earlier, I built developer-facing tooling and high-velocity UIs at Oracle Cloud Infrastructure and Amazon. More recently I've been working at the edge of AI-assisted web platforms — AI observability, evaluation tooling, and developer-productivity workflows — bringing the same rigor to AI-powered experiences that I bring to traditional frontend systems. I hold an M.S. in Information Sciences and Technologies from the Rochester Institute of Technology (4.0 GPA).
Experience
Frontend & full stack engineering across eBay, Oracle Cloud, and Amazon
-
Member of Technical Staff 2 · eBay
Staff-scope frontend engineer on eBay's highest-traffic commerce surfaces, owning frontend architecture, web performance, and observability standards across React, Marko, TypeScript, Node.js, and Java.
- Architected a vendor-agnostic, real-time chat platform across eBay Live, Support, and Auctions supporting 2M+ concurrent users at sub-200 ms latency, with a transport-agnostic adapter reused by React, Marko, and Native teams.
- Built a lightweight OpenTelemetry frontend observability framework adopted by 100+ eBay applications (~99% of frontend apps), contributing to a 15% reduction in bugs and alerts.
- Diagnosed stale auth-token failures surfaced by that framework, cutting Add-to-Cart errors from 2% to 0.1% — an estimated $1M+ in annual GMV.
- Optimized Buyer Experience pages serving 10M+ daily views, cutting page load by 80 ms and improving INP by 12%; earned a Spot Award.
- Defined the "One Click Away" integration blueprint for third-party View Item widgets, standardizing SSR above the fold and micro-frame modal patterns across Buyer Experiences.
- Built an upstream seller-content validation service that combined rule-based HTML checks with AI-generated fix explanations, cutting seller-description incidents ~80% and engineering escalations 70%+.
JavaScript · TypeScript · Java · Node.js · Marko · React · GraphQL · OpenTelemetry
-
Senior Member of Technical Staff · Oracle Cloud Infrastructure
Built and modernized React frontend layers and backend services for cloud-security tooling, improving both interactivity and read-path latency across the console.
- Designed modular React UI layers for security tooling, using lazy loading and code splitting to reduce Time to Interactive by 15% across the console.
- Built Go and Java APIs backed by Redis caching layers that cut P90 read-path latency from 900 ms to sub-300 ms.
- Modernized React and Redux patterns and shared tooling across Security Central, improving delivery consistency and reducing repeated implementation work across feature teams.
Java · TypeScript · SQL · React · Redux
-
Front End Engineer · Amazon
Built reusable UI component libraries and guided console workflows for high-velocity Amazon surfaces and AWS developer tooling.
- Built reusable UI component libraries for high-velocity pages across Amazon.com and Deals; refactored rendering execution paths to reduce client-side CPU latency by ~20%.
- Designed and drove guided UI workflows for the ECS Console by integrating disparate service APIs, growing customer usage and call volume 40% in one year.
- Engineered infrastructure-automation pipelines that reduced service-region provisioning and validation cycle times from 3 weeks to 2 days.
JavaScript · TypeScript · React · Redux · Angular · Webpack · Docker · AWS
-
Co-Op JavaScript Developer · Oracle (Primavera GBU)
Implemented and tested new UI abstractions with Chrome DevTools, refactoring legacy code and raising unit-test coverage by 10%.
-
System Engineer · Infosys Limited
Built hybrid mobile apps for Pfizer with Cordova and developed reusable UI frameworks that cut delivery time by 20%.
Education: M.S., Information Sciences and Technologies — Rochester Institute of Technology (GPA 4.0/4.0) · B.Tech, Computer Science and Engineering — Jaypee Institute of Information Technology.
Skills
A frontend-first toolkit, backed by full stack and observability depth
Backend / Full Stack
- Node.js
- GraphQL
- REST APIs
- Java
- SQL
- Real-time systems
Observability
- OpenTelemetry
- Distributed tracing
- Logging
- Metrics
- Session replay
- Production debugging
Testing / Quality
- Playwright
- Percy (visual regression)
- Automated testing
- Production readiness
- Core Web Vitals / INP
AI / Developer Productivity
- AI-assisted workflows
- AI observability
- Evaluation (evals) tooling
- Model Context Protocol (MCP)
- Claude API · OpenAI API
- Vercel AI SDK
Cloud / Infrastructure
- AWS (ECS, CloudFormation)
- Docker
- CI/CD
- Caching layers
Featured Work & Impact
Projects where frontend architecture, observability, and performance met real business impact
Real-time chat platform at scale
eBay · Buyer Experiences
Architected a vendor-agnostic, framework-agnostic real-time chat platform spanning eBay Live, Support, and Auctions. A transport-agnostic Chat Adapter Contract let React, Marko, and Native teams share one core, with virtualized rendering and optimistic UI keeping the experience smooth under load.
- 2M+ concurrent users
- <200 ms P50 round-trip latency
- Zero frontend rewrites to swap vendors
Frontend observability & production readiness
eBay · Platform / Buyer Experiences
Built a lightweight OpenTelemetry frontend observability framework into the app shell as an opt-out standard, then defined production-readiness evals so instrumentation became part of the definition of done. The signal it surfaced turned invisible failures into measurable fixes.
- 100+ apps adopted (~99%)
- 15% fewer bugs & alerts
- Add-to-Cart errors 2% → 0.1% (~$1M+ GMV)
High-traffic commerce & Core Web Vitals
eBay · View Item
Protected the critical rendering path on pages serving 10M+ daily views — reducing blocking JavaScript, improving hydration, and standardizing the "One Click Away" pattern (SSR above the fold, micro-frame for modals) so third-party widgets couldn't regress performance or SEO.
- −80 ms page load
- +12% INP
- Spot Award for performance impact
Shared frontend platform & developer experience
eBay · Frontend Platform
Built Playwright-based testing utilities and Percy visual-regression coverage adopted broadly across eBay frontend teams, and modernized shared React/Redux tooling — strengthening pre-merge quality while reducing onboarding friction for new engineers.
- −40% onboarding friction
- Broad cross-team adoption
- Stronger pre-merge quality gates
AI observability assistant
eBay · AI & Observability
Built an AI assistant that correlates session replay, distributed traces, and logs, then answers natural-language root-cause questions over MCP. Evaluation frameworks measured answer quality and trace accuracy against real production incidents.
- 20–30% faster time-to-diagnosis
- 20–30 min saved per investigation
- Self-service production debugging
AI-assisted developer tooling & agents
eBay tooling · Self-built projects
Built Claude Skills that flag visual, rendering, and bundle regressions from npm changes inside the developer inner loop (locally and as a GitHub Action). Separately, I built a terminal AI agent from first principles in TypeScript — a full tool-calling loop with context compaction, human-in-the-loop approval gates, and LLM observability — as a personal project.
- Pre-CI regression detection
- Tool-calling agent loop & HITL gates
- Evals & LLM tracing
Contact
Let's talk frontend, full stack, and observability
I'm always glad to connect with teams working on ambitious frontend platforms, web performance, observability, or AI-assisted developer experiences. LinkedIn is the best way to reach me.