Aryan Garg
Full-Stack Engineer
Building Products that ship. Frontend-Heavy, Full-Stack at Heart.
- Role
- Software Engineer 1 (Full-Stack, Frontend-heavy) @ Deloitte Consulting USI
- Location
- India
- Local time
- --:--:-- IST
- Resume
- View CV →
- Pronouns
- He/Him/His
About
Full-stack engineer building production web apps end-to-end — strongest on the frontend (React, Next.js, TypeScript, design systems, performance), and comfortable across the stack with Node.js, Express, PostgreSQL, MongoDB, and real-time APIs. I love bridging design and engineering — turning complex interactions into smooth, intuitive interfaces while shaping the APIs and data models that power them. Currently working on AI-powered developer tools and creative web experiments outside of work.
- Full-Stack Developer with a passion for building high-performance, user-centric software solutions with intuitive and engaging designs.
- Skilled in Next.js, React, TypeScript, and modern front-end technologies; building high-quality, user-centric web and mobile applications.
- Enthusiastic about exploring new technologies and turning ideas into reality through polished, thoughtfully crafted personal projects — currently Feedforward, TableSense, and GatherGo.
Contributions
Stack(43)
Experience
Deloitte Consulting USI
Software Engineer 1
- Employment Type
- Full-time
- Employment Period
- 01.2024—Present
Building enterprise web applications for global consulting engagements. Focused on the frontend — React, JavaScript, Design System Work, Documentation and Performance — while owning slices of the backend in Node.js, Express.js and working closely with databases (PostgreSQL, MongoDB) and real-time APIs along with the Cloud Platforms such as AWS.
- Built and shipped React-based dashboards and feature modules for enterprise consulting engagements, prioritizing performance, accessibility, and consistency across a shared design system.
- Owned end-to-end frontend development on multiple modules — translating Figma designs into responsive React + JavaScript components, with reusable patterns adopted by other teams.
- Contributed to and extended an internal component library used across multiple projects, standardizing form, table, and dashboard primitives to cut feature delivery time.
- Improved page-load and runtime performance on a high-traffic internal dashboard via code-splitting, memoization, lazy loading, and image optimization.
- Worked across the full stack — built REST APIs in Node.js / Express alongside React frontends, handling auth, validation, error handling, and pagination end-to-end.
- Designed and integrated REST APIs between React frontends and backend services, owning the contract from schema.
- Prototyped and shipped GenAI-assisted features — LLM-powered search, summarization, and form-fill helpers — integrating OpenAI and Anthropic APIs with proper rate limiting, retries, and prompt-injection guardrails.
- Built agentic AI workflows for internal tooling — multi-step LLM agents with tool use, structured outputs, and fallbacks — reducing manual review time on a recurring workflow.
- Worked in distributed Agile teams across time zones — leading sprint planning, writing technical design docs, and conducting code reviews to keep velocity high without sacrificing quality.
- Authored architecture and onboarding documentation in Confluence; maintained Jira boards as the source of truth for sprint scope and blocker tracking.
- Mentored junior engineers on React patterns, and frontend performance — pairing on tickets and reviewing PRs as a regular part of the role.
Education
Dr. B. R. Ambedkar National Institute of Technology, Jalandhar
B.Tech, Information Technology
- Period
- 07.2019—06.2023
- Grade
- CGPA 8.43 / 10.0
Budha Dal Public School, Patiala
Class XII (CBSE), Senior Secondary
- Period
- 04.2017—04.2019
- Grade
- 85%
Work(3)
Feedforward
06.2026 — Present
Embeddable customer-feedback platform (NPS/CSAT/text surveys) with AI insight extraction. The engineering showcase is a dual-bundle frontend: a <30KB Preact widget that runs inside arbitrary third-party host sites via closed Shadow DOM, paired with a React 19 dashboard streaming live updates over SSE.
Featured
- Sub-30KB embeddable widget — Preact + Vite IIFE library mode, closed Shadow DOM isolation, CI hard-fails over the 30KB gzipped budget
- Two-trust-domain security model — HMAC-signed embed tokens (replay-protected via Redis nonces, timingSafeEqual) for untrusted host sites; Clerk bearer auth for the dashboard
- Three distinct AI integration patterns — async BullMQ batch (theme/sentiment), on-demand sync (question improvement), and scheduled two-stage anomaly detection (z-score pre-filter → LLM)
- Live dashboard over SSE — per-org Redis pub/sub channels, 25s heartbeats, auto-reconnect with polling fallback
- Multi-tenant isolation — orgId always derived from token, never request body; per-org Redis channels
- Stripe usage-based billing — Redis INCR metering, hourly flush to Stripe, idempotent webhooks
TableSense
06.2026 — Present
A visual floor-plan & live-operations platform for independent restaurants. Design your dining room on a canvas, then run service in real time with multi-device sync, reservations, and offline-first PWA support.
Featured
- Konva canvas floor-plan editor — drag-and-drop table layout, versioned design-time positions kept separate from live runtime state
- Socket.IO real-time multi-user sync — last-write-wins conflict resolution on a per-table version counter (verified 29ms two-device latency), fanned out via Upstash Redis
- Offline-first PWA using all four browser storage layers, with a custom Workbox service worker (injectManifest) owning the Web Push handler
- Dual-channel notifications — guaranteed email (Resend) + fire-and-forget Web Push (VAPID)
- Append-only audit log drives analytics — turn time, covers-per-hour, and no-show rate computed from TableEvent history, not live aggregates
- Multi-tenant security — restaurantId scoped from JWT claims, Stripe webhooks idempotent and raw-body verified
- Self-hosted logging dashboard with PII-scrubbing client logger
GatherGo
06.2026 — Present
A multi-vendor event ticketing platform. Organizers publish event pages, sell tiered tickets via Stripe, issue QR-coded tickets, and check attendees in from a mobile door app. Ticketing for independent organizers, without the Eventbrite tax.
Featured
- Oversell-proof checkout — Stripe Checkout backed by per-tier Redis locks + an idempotent webhook, so the last ticket never double-sells
- Signed QR tickets (JWT) + an installable check-in PWA with atomic, double-scan-proof check-in
- Plan-gated AI (Groq) — event descriptions, category/tag suggestions, and plain-English sales insights, locked to paid tiers
- Stripe subscription billing (Free/Pro/Business) with a self-serve membership page
- Live organizer analytics (Recharts, polling) + CSV attendee export
- Production auth with Clerk (Google sign-in) + webhook user-sync; abandoned-cart sweep cron
- Calm, accessible "Sense"-inspired UI: shimmer skeletons, interactive hero bento, keyboard shortcuts, light/dark
Blog(6)
- Node.js in 2025: Modern Practices You Should Be UsingFeaturedLet's Code Future
- 10 CSS Hacks Every Frontend Developer Should Know in 2025FeaturedStackademic
- The Interface Is Alive: How AI Is Quietly Rewriting the Rules of UI DesignFeaturedDesign Bootcamp
- The UI That Builds Itself: How Generative AI Is Rewriting the Rules of Interface Design14.06.2026When the interface stops being a blueprint and starts being a conversation, everything changes. Continue reading on Bootcamp »…
- Bun Just Rewrote a Million Lines of Code in 6 Days — and an LLM Did Most of It06.06.2026What Bun’s Zig-to-Rust migration tells us about the future of codebases, memory safety, and why “done by AI” is not the same as “done… Conti…
- Your Interface Is Screaming. It’s Time to Make It Whisper.30.05.2026Calm UX and the Anti-Maximalism Movement Are Quietly Rewriting the Rules of Digital Design — and the Users Are Already Grateful. Continue re…
AI Concierge
Three tools powered by Groq. Pick one.
Powered by Groq · llama-3.3-70b
