Feedback Hub is a full-stack feedback and survey SaaS for building forms with a drag-and-drop builder and conditional logic, sharing them through public links, QR codes, or tracked campaign links, and turning responses into funnel, NPS, sentiment, and segment analytics. NPS-triggered follow-up emails with open tracking, Slack project notifications, custom branding, an audit log, multi-workspace roles, and Stripe billing run on a Next.js frontend backed by a dedicated Fastify + tRPC API over PostgreSQL.

A team that collects feedback usually runs five tools at once: a form builder, a spreadsheet for responses, an analytics product, an email tool for follow-ups, and a billing system. Nothing shares a workspace, roles, or a single place to act on what a respondent said — so the loop from "ask" to "follow up" never closes in one place.
Feedback Hub is a full-stack feedback SaaS that owns the whole loop: a drag-and-drop form builder, public sharing by link or QR code, response collection, an analytics layer (funnels, NPS, segments), and follow-up emails to the people who answered — inside multi-workspace teams with role-based permissions and Stripe subscription billing.
A Next.js App Router frontend talks to a dedicated Fastify 5 + tRPC backend over PostgreSQL (Drizzle). The frontend imports the backend only as a type, so the API contract is shared end to end with no server code in the browser bundle, and both sides validate with the same Zod schemas.
| Layer | Technology | Where it runs |
|---|---|---|
| Frontend | Next.js App Router | Vercel |
| API | Fastify 5 + tRPC (Node.js) | PM2 behind nginx on an AWS EC2 instance |
| Database | PostgreSQL via Drizzle | AWS RDS |
| Sessions & jobs | Redis (refresh sessions, BullMQ queues) | reachable via REDIS_URL; feature-gated |
| Media | S3 object storage | Amazon S3 (MinIO locally) |
| SMTP | provider via SMTP_* (Mailpit locally) | |
| Billing | Stripe | Stripe API + signed webhook |
The frontend and backend deploy independently — a Vercel build for the web app, a PM2 restart from a self-contained bundle on EC2 — and share nothing but the tRPC type contract. Every service after the database is feature-gated: Redis, SMTP, S3, Stripe, OAuth and PostHog each stay off while their env vars are empty, so the API boots on DATABASE_URL alone.
flowchart TD
subgraph Web["Next.js App Router · Vercel"]
Builder[Form builder & dashboards]
Public[Public form · /form/share token]
end
Web -->|type-safe tRPC over HTTP, cookies| API
subgraph Server["Fastify 5 · PM2 + nginx · AWS EC2"]
API["@org/backend baseline<br/>request-id · helmet · CORS · /health"]
API --> Router[tRPC appRouter]
Router --> Svc[Service · one folder per use case]
Svc --> Repo[Repository · Drizzle queries]
Hook["POST /webhooks/stripe<br/>raw body · signature verified"] --> Svc
end
Repo --> DB[(PostgreSQL · AWS RDS)]
Svc -.sessions + BullMQ jobs.-> Redis[(Redis)]
Svc -.follow-up email.-> SMTP[SMTP]
Svc -.media.-> S3[(Amazon S3)]
Stripe[Stripe] --> Hook
Stripe is the one inbound exception to the tRPC path: its webhook runs in its own Fastify plugin scope with a raw-buffer body parser, so the signature is checked against the exact bytes while /trpc keeps receiving parsed JSON on the same server.
sequenceDiagram participant B as Browser participant V as Next.js · Vercel participant F as Fastify · EC2 participant S as Service + Drizzle participant D as PostgreSQL · RDS B->>V: open page / submit form V->>F: tRPC call /trpc/router.procedure (HttpOnly cookies) F->>F: baseline — request id, helmet, CORS F->>F: procedure — validate Zod input, check auth level F->>S: call the use case S->>D: Drizzle query (one pooled connection) D-->>S: rows S-->>F: result F-->>V: typed JSON (or TRPCError via toTrpcError) V-->>B: rendered UI
A procedure validates its Zod input and auth level before any service runs, so a bad request never reaches the database. On UNAUTHORIZED the web client calls auth.refreshSession once and retries, so an expired access token is invisible to the user. Write paths can hand work off asynchronously — submitting a response enqueues a BullMQ job instead of blocking the responder — and errors cross the boundary as a typed TRPCError (AppError → toTrpcError), never a raw stack trace.
@org/backend baseline (request id, helmet, CORS, health probes) comes for free.drizzle-kit. One pooled connection is cached per process.REDIS_URL the API still runs, only without those paths. Infrastructure is a choice the deployment makes, not a hard dependency.stripe-signature header verifies against the raw body, never from a value the client sent.Sole full-stack engineer. Designed the service → repository layering, the type-only tRPC contract between the two apps, the isolated raw-body scope for the Stripe webhook, per-workspace access checks on every procedure, and the review-before-send follow-up pipeline. The frontend deploys to Vercel; the backend runs on PM2 behind nginx on an EC2 instance, against PostgreSQL on AWS RDS.
One Fastify server had to verify a Stripe webhook signature — which needs the untouched raw request bytes — while every tRPC route needed parsed JSON. Parsing the body globally breaks the signature; not parsing it breaks everything else. The fix was to mount the webhook in its own plugin scope with a raw-buffer parser, leaving the global JSON parser in place for /trpc, so both contracts hold on the same process.

Real Estate Media Pricing Calculator is a web-based pricing platform that helps real estate media professionals generate accurate, market-based pricing recommendations for photography, videography, drone services, floor plans, virtual staging, and twilight photography across the United States. The platform uses verified market data from 2,695 real estate media businesses to provide location-specific pricing ranges with confidence ratings, helping professionals price services competitively and confidently.

Extra Scrub Cleaning Services Website is a professional multi-page business website built for a cleaning company based in London. The platform helps customers explore residential and commercial cleaning services, request custom quotes, book appointments online, and contact the team easily. The goal was to create a modern, mobile-friendly digital presence that improves customer trust, simplifies service discovery, and increases booking conversions.

Urban DTF Customer Portal is a full-stack web application built for URBAN DTF & T-SHIRTS, a custom DTF printing business. The platform allows customers to upload designs, configure garment orders, make secure payments, and track production progress in real time. It helps streamline order management, reduce manual communication, and improve customer experience by transforming traditional printing operations into a fully digital workflow.