I build and scale production web platforms with React, Next.js, and TypeScript — from architecture and system design to implementation, performance, and production delivery.
Recently extended into AI-native engineering — shipping a streaming LLM assistant and the application layer for a multi-tenant RAG platform.
Lead Software Engineer and Frontend Architect with 10+ years of experience designing, building, and scaling high-performance web platforms across consumer, B2B SaaS, e-commerce, food tech, and market intelligence products. I currently lead a 9-person frontend engineering team at CoffeeWeb, owning frontend architecture for a mobile-first PWA serving 200,000+ users across 175+ countries — including re-architecting an immature, hard-to-maintain codebase into a structured system with clear component boundaries, consistent state/data patterns, and production observability. Before that, I led frontend engineering at Peppo across BookMyShow Deals and a suite of B2B consoles, helped build JFrog Pipelines from inception as part of the JFrog platform — its Vue.js UI and the Node.js client behind it — helped scale Shippable's CI/CD platform to 100,000+ Docker containers a month, and built CenturyLink's e-commerce and mobile ordering platforms. Across every team, I've owned tech stack decisions, hiring, and production reliability — and I'm now extending that decade of frontend architecture into AI-native engineering: I've shipped a streaming LLM assistant (Ask AI) and built the application layer for YourBot, a multi-tenant RAG chatbot platform. This entire site — architecture, the live streaming Ask AI assistant, and deployment — was designed, built, and shipped by me alone, working remotely, with no team in the room.
Want the short version? Ask the AI assistant trained on my experience.Ask AI
Worked with teams at
A compact view of the tools I use most often to ship products.
Tech stack
How I work
AI coding tools
CoffeeWeb Technologies · Remote
Privately held market intelligence and technology platform headquartered in Bengaluru — the definitive global hub for information, supply chain data, and digital resources in the coffee sector, connecting 200,000+ users across 175+ countries.
Project: CoffeeWeb Platform, CoffeeWeb Admin
Peppo Technologies · Remote
MarTech company helping brands with customer acquisition and retention — products include the headless loyalty infrastructure RewardX, an online ordering system for restaurants and cloud kitchens, and a WhatsApp-delivered e-invoicing solution.
Project: BookMyShow Deals, Peppo PWA (white-label online ordering), Business Console (Merchant/DMS/RewardX), Event food ordering (Lollapalooza India, Backstreet Boys DNA World Tour)
JFrog India · Bengaluru
On a mission to enable continuous updates through Liquid Software. The world's top brands — including Amazon, Facebook, Google, Netflix, Uber, VMware, and Spotify — are among the 4,500+ companies that depend on JFrog to manage binaries for mission-critical applications.
Project: JFrog Pipelines — UI module in JFrog Platform UI, Pipelines Node.js client
Shippable India · Bengaluru
Automates CI/CD and DevOps activities with streamlined workflows across teams and tools — available as an on-premises server product, hosted SaaS, and a hybrid offering called Custom Nodes.
Project: Shippable Platform — web app: job dashboards, node pools, subscriptions & billing, admin console
CenturyLink India · Bengaluru
Global communications and IT services company focused on connecting its customers to the power of the digital world.
Project: E-Commerce, Instalink
C-DAC ACTS, Bangalore
Autonomous - Ministry of Electronics and Information Technology (MeitY), Government of India.
Oriental College of Technology, Bhopal
Rajiv Gandhi Proudyogiki Vishwavidyalaya (RGPV), Bhopal
Campion School, Bhopal
CBSE
Campion School, Bhopal
CBSE
Ask AI and YourBot are two different AI builds, not the same trick twice — one shows LLM integration, streaming, and context engineering, the other shows RAG architecture and multi-tenancy.
A live, streaming chat assistant (this site's /ask-ai page) that answers questions about my experience. Grounded using context stuffing — the entire knowledge base (experience, skills, public log entries) is assembled server-side and injected directly into the system prompt on every request, a deliberate choice over RAG since the corpus comfortably fits in the model's context window with no retrieval step needed. Built on the Gemini API, streams tokens to the client over a ReadableStream, enforces per-IP rate limiting, and strips private/draft content server-side before it ever reaches the model.
Step 01 · You ask
Your question goes to the server along with the last few turns of the conversation, so follow-ups like “and before that?” still make sense.
A multi-tenant SaaS platform for building, configuring, and embedding custom RAG chatbots trained on a business's own documents. I own the application layer end to end — tenant/team management, the Next.js bot-creation and configuration UI, document management, and the integration surface into backend AI services — while the LLM inference and retrieval execution itself is handled by a separate backend AI service this app integrates with.
Ingest a document · 01 Tenant & auth
Built by meSigning up creates the company's account in Postgres and its user in Keycloak through the Admin API, sends the verify-email action, and assigns the free plan. Access and refresh tokens live in httpOnly cookies, JWTs are verified against Keycloak's JWKS, and sessions refresh silently.
solid tile: I built it dashed: backend AI service I integrate with
My ownership
Backend / AI service
A B2B back-office console for restaurant and venue operators — sales dashboards, restaurant/venue onboarding, an item library with categories, modifiers and sales-tax rules, role-based user management, and table management. Rebuilt the entire frontend from the ground up as a migration off the platform's legacy Angular UI to React, using Material UI and react-admin for the admin interface, React Router for navigation, and Axios with JWT bearer-token auth against a Spring Boot backend.
A personal finance PWA for tracking multiple financial identities under one account — banks, credit cards, loans, investments, and recurring bills, each scoped to its own profile. Built as a Next.js App Router PWA on Supabase (Postgres + Auth), with a data model built around per-user "profiles" (personal, professional, family, spouse — up to five) rather than a single flat account, so every bank, credit card, loan, investment, recurring payment, and transaction is scoped to a profile and isolated via Row Level Security policies that join back to the owning user through auth.uid(). Session handling is hand-rolled rather than the standard Supabase SSR helper — access/refresh tokens are set as httpOnly cookies on login and re-attached per request in middleware, which gates every dashboard route. Feature set spans credit card utilization tracking (limit/utilized/available), EMI loan tracking with part-payments and remaining balance, friend-and-family lending with status and repayment schedules, recurring bill reminders, and an investments ledger — with recent work moving bank/card data fetching into Server Actions and adding page-level cache control for fresher dashboard reads.
I turn product requirements into technical direction, build systems that scale, help teams execute, and stay accountable for what happens in production.
Leading a team hasn't taken me off the keyboard. I still own hard problems end to end: I built CoffeeWeb's Node.js market-data service (external data collection, currency processing, event publishing), the application layer for YourBot (a multi-tenant RAG platform), and a full Angular-to-React rebuild of a venue-management console. I scope the work, unblock myself, write it up clearly, and ship without needing someone to check in.
I design frontend systems around real product requirements, not frameworks or trends. At CoffeeWeb, I architected a mobile-first PWA spanning 16 product domains for 200,000+ users across 175+ countries, and re-architected the existing codebase into clear component boundaries, consistent state and data patterns, and production observability.
I lead a 9-person frontend team while staying roughly 60% hands-on — making architecture calls, writing and reviewing code, and setting engineering standards. I own hiring, mentoring, and performance reviews, and build teams that make good decisions without me in the room.
I take work from PRD to release: shaping the technical solution, breaking roadmaps into milestones, and coordinating with product, design, and backend. Shipping isn't the finish line — at CoffeeWeb I set up the quality foundation: ~165 Jest/React Testing Library test files, Sentry monitoring, and Azure DevOps CI/CD across dev, staging, and production.
Fully remote since 2020, including leading two frontend teams — at Peppo and CoffeeWeb — from hiring through delivery. I document decisions instead of relying on hallway context, run PR-based reviews and async standups, and stay accountable to outcomes, not hours online.
I built the application layer for YourBot, a multi-tenant RAG platform (pgvector retrieval, document ingestion, Keycloak RBAC), and shipped this site's streaming Ask AI assistant on the Gemini API. I'm also leading how my team adopts AI-assisted development.
Broad perspective. Shaped by work across consumer, B2B SaaS, e-commerce, food tech, DevOps, and market intelligence products — so I weigh frontend decisions against the larger system and the business.
It answers from my experience, projects, and writing — the same knowledge base this site is built from — and streams the answer in seconds.
Open to senior and architect-level frontend roles, AI-adjacent full-stack work, and consulting in platform engineering, technical leadership, and system architecture.
Roles
Senior / architect-level frontend, and AI-adjacent full-stack.
Consulting
Advisory, audits, or hands-on — architecture, platforms, and team leadership.
Based in
Bhopal, India · IST (UTC+5:30) · fully remote since 2020.