I architect a framework-agnostic design system consumed by Angular, React, Vue and vanilla JS teams — and in parallel I take products from an empty repository to a deployed system: schema, API, interface, native builds, tests, CI.
The cost of writing code collapsed. Scaffolding, tests, migrations and documentation are no longer where the hours go.
Knowing what to build, how to model it, where it breaks at scale, and whether it is accessible, tested and maintainable.
Thirteen years of architecture judgement, driving AI at full speed. Output that is designed — not merely generated.
HR, attendance, leave and payroll; projects with Kanban, timesheets, purchase and inventory; sales & CRM pipeline, finance and expenses; an approvals engine, reporting and audit — behind role-based access control and a full REST API, with face-recognition attendance capture.
Built against a transcribed enterprise SRS the document itself scoped at 14–18 months for a full team: a web ERP for every staff role, plus Patient and Employee PWA apps on one API. Argon2 auth, RBAC, structured logging, and a role-by-role smoke test — with status tracked honestly against the spec.
Customer mobile app, partner app, admin and ops panel, and a realtime backend — delivered together in a single monorepo, with Android APKs packaged through Capacitor and a designed premium visual system rather than a default theme.
React Server Components querying PostgreSQL through Prisma at request time, packaged as an installable Android application via Capacitor. Verified pipeline — typecheck, 83 unit tests, lint — plus a five-role route smoke test run against both the dev server and a production build.
Runs a whole group — manpower and payroll, printing, signage and fabrication — from first enquiry through survey, quotation, approval, advance, production, QC, delivery and installation, plus deployment, attendance, live tracking, HR, payroll and inventory. One /api/* backend serves the web app and the Android client, so what a supervisor captures on a phone lands in the same system.
Drizzle ORM over Neon serverless PostgreSQL, a Radix and shadcn component layer, TanStack Query for the data layer, and blob storage for clinical documents — an EMR built to be operated, not demoed.
A Go 1.24 desktop application with an embedded web frontend, shipped as an NSIS Windows installer so it deploys onto machines with no runtime prerequisites — the whole product in a single executable a shopkeeper can double-click.
Fleet management down to fuel, maintenance, inspections and parts; live vehicle tracking with trips, playback, geofences and alerts; and a logistics ERP covering quotations, bookings and consignments — six licensable modules sharing one login, one organisation and one master-data set, with dedicated driver and agent Android apps.
The whole property lifecycle on one connected spine — marketing and listings, lead capture and sales, construction execution, unit booking and collections, handover, then post-sale rental and facility operations. Built to an SRS with six end-to-end flows traced from requirement to route, and shipped with its architecture and traceability documents.
Manufacturing Solutions Management System, authored as an SRS before a line was written — module design, process flows, a KPI catalogue and a phased roadmap — then implemented across separate server and web workspaces so the two can scale and deploy independently.
Project, resource and site-operations software for construction firms — specified as a written requirement document first, then delivered as a working MVP on Next.js with Prisma and TanStack Query rather than a deck full of screens nobody can click.
Stock control delivered across three surfaces at once — a web application, an Android client and Odoo add-ons over shared infrastructure — built one traced feature branch at a time: authentication, goods receipt, then the movement ledger, each landing as its own reviewable slice.
An attendance client written natively in Java rather than wrapped from the web — project switching, filtered reporting and offline-tolerant sync, because the field teams using it are frequently somewhere with no signal and the record still has to be right when they get back.
Tracks a family chit the whole way through — monthly instalments, who took the chit and when, extras paid afterwards, carry-over months, requests, and the closing settlement for every member. Written in Kotlin and Jetpack Compose on Supabase, in English and Telugu, because half the people using it read one and half the other.
A static marketing site rebuilt as a server-rendered application on Next.js 16 and React 19 — all eighteen pages reproduced one-to-one in design, content and behaviour, including product detail, free-trial, demo-booking and subscription flows, but now crawlable and fast on first paint.
LLM applications, retrieval-augmented generation over real data, agentic tool-calling workflows and MCP servers — built with evals and guardrails rather than vibes.
Framework-agnostic component libraries on Web Components, distributed as native Angular, React and Vue packages from one codebase — tokens, accessibility, Storybook, visual-regression gates.
Angular 7 through latest with Standalone APIs and Signals, React 19, Next.js 15 App Router and RSC, micro-frontends, and performance work that shows up in the numbers.
Schema design in Prisma and Drizzle over PostgreSQL, REST services in Node, .NET 9 and Go, auth and RBAC, multi-tenancy including database-per-tenant, realtime over Socket.io.
The same product surface shipped as an Android APK, a Windows desktop application and a PWA — Capacitor, Flutter, Tauri, Electron and Go, with real installers.
Vercel, Supabase, Neon, Render and Docker; GitHub Actions pipelines with typecheck, unit tests, lint and smoke tests as gates. Coverage written with AI, reviewed by a human.
Architecting a framework-agnostic design system on StencilJS and native Web Components — one codebase consumed by Angular, React, Vue and vanilla JS teams. Built the distribution layer including all three Angular output strategies (NgModule, SCAM, Standalone). Published packages to the private Figma npm registry so designers generate production-consistent code from real components. Chromatic visual regression and Storybook interaction tests run as pre-merge CI gates.
Built complex Angular and DevExpress screens against microservices for a global credit-risk platform, with interactive visualisations powering quantitative risk dashboards. Led architecture of a reusable Angular component library consumed across multiple business applications.
Developed reusable Angular and Angular Material components for the open-source ArkCase platform; implemented Jest regression suites and managed JSON Server mock APIs ahead of backend delivery.
Built the front end with scalable Angular component architecture and REST integration; developed custom directives and reusable service layers for a global electronics repair and service platform.
Led front-end development of an enterprise commercial banking platform, architected a reusable component library and REST service layer for banking workflows, and led client demos through Agile sprints.
Owned the UI lifecycle for the majority of customer-facing pages from design through to production on India's leading digital healthcare platform, and integrated analytics and growth tracking across the funnel.
Built responsive home, auto and funeral insurance platforms for South Africa with full cross-browser compatibility, plus responsive email templates and campaign landing pages.
Developed the front ends for chit-fund industry management and student life-cycle applications, including internationalisation, input validation and cross-browser support.
Every build starts with an SRS or architecture document — modules, process flows, data model, KPIs, phasing. AI drafts it fast; the decisions in it are mine.
Prisma or Drizzle schemas, relationships, tenancy strategy and access rules before a single screen exists. Most product failures are data-model failures.
AI produces components, services, migrations and tests at pace. I review every boundary — auth, money, medical data, accessibility — because that is where it costs.
Typecheck, unit tests, lint and route smoke tests run before merge. Repositories ship with enforced branching rules and honest status tracking, not optimistic READMEs.
Ravi Kumar Medam
Full-Stack Engineer · AI Architect
