SWATI PORWAL
Projects

Things I've shipped.

Enterprise B2B platforms and B2C products from inside Kazam, and AI-native products I've built solo, end to end.

Tap or hover a card to expand the full case study

Kazam billing integrity
B2B SaaS

Trust at Scale

Real-time billing integrity engine for 10+ enterprise clients across 50+ network operators.

10+enterprise 30–50%fewer disputes 2–3h→0resolution
Expand case study
Shipped · Kazam EV Tech

Trust at Scale: Making Every kWh Auditable

A CMS-level auditor that cross-verifies every charging session against charger telemetry, payment records, and timestamps — auto-correcting anomalies before they reach a customer's bill.

"Data needs to be sacrosanct. If your chargers are sending incorrect data, your invoicing, auditing, and customer trust are all up for a toss."
Problem
Mismatched charger logs silently broke billing and client trust.
System
Always-on auditor reconciling firmware, CMS, and payment data in real time.
Outcome
Manual resolution (2–3 hrs/case) dropped to near-zero; anomaly support volume down ~50%; protected the ~5–10% of revenue leaking into disputes.
Enterprise SaaSBillingReal-time
Full case study ↗ Read the article ↗
B2B2C · Mobile

One Core, Fifty Brands

Ground-up B2B2C app revamp — 45–50 white-label builds, EMEA & SEA. Auto-charge, trip planning, configurable core.

20%+fewer UX tickets 0.5s→msload time 45–50branded apps
Expand case study
Shipped · EMEA & SEA · iOS & Android

One Core, Fifty Brands

Ground-up revamp of a B2B2C consumer app for an EV charging network — the same core powers 45–50 separately branded App Store and Play Store builds. I owned discovery through launch with a squad of 2 designers, 3 engineers, 2 QA, and 1 support.

Problem
Inconsistent UI, ~0.5s loads at the charger, too many steps to start a session — and every fix required updating all 45–50 client builds through store review.
Solution
Full flow redesign, auto-charge, trip planning with reliability signals, cost-before-charge, and a configurable plug-and-play core for white-label deployments.
Outcome
20%+ fewer UX support tickets, load time to milliseconds on key interactions, ~5% lower crash rate, enterprise escalations on app quality stopped.
B2B2CWhite-labelAuto-chargeTrip planningiOS & Android
Full case study ↗
Soul Journal app
Solo build

Soul Journal

Privacy-first emotion journaling — no account, entries stay on your device.

0accounts 6core feelings 100%solo build
Expand case study
Live · On-device entries

Soul Journal

A two-layer emotion wheel for naming what you actually feel. Entries live only on your device — privacy as a constraint, not a feature.

Problem
Journaling tools ask you to trust a server with your most private thoughts.
Approach
No account, no entry database — journals persist locally on the device.
How
Six core feelings, each with granular nuance mapped beneath.
GeminiAI StudioOpen source
Open the app ↗ GitHub ↗
Fridge Chef home screen
Live product

Fridge Chef

Ingredient-first Indian meal companion — pantry to recipe, weekly macros, WhatsApp cook handoff.

183pantry items 6macro presets 4diet modes
Expand case study
Live · Next.js · Vercel

Fridge Chef

A mobile-first meal companion that turns pantry ingredients into diet-aware Indian recipes — with macro tracking, weekly planning, and WhatsApp handoff to household cooks. Select what's in the fridge, get ranked matches, plan the week, compare kcal and protein to goals, share to WhatsApp, and order missing items via quick-commerce.

The problem
Recipe apps start with content, not context. Indian households need meals from what's already at home — respecting veg/vegan rules, allergies, macro goals, and cooks who work from WhatsApp instructions.
The approach
Ingredient-first flow with a curated DB and multi-factor scoring on match, cuisine, health tags, allergens, and macro fit — instant, trusted results without AI dependency. Macro goals surfaced on the home screen at decision time.
The build
5-tab core journey: select → match → plan → execute. WhatsApp share + Blinkit/Swiggy Instamart/Zepto deep links. Next.js 16, React 19, zero-login profiles, deployed on Vercel.
Next.js 16React 19Indian recipesMacro trackingWhatsAppVercel
Open the app ↗ Product one-pager ↗ GitHub ↗
TaskQuest login and dashboard
Full-stack

TaskQuest Tracker

Production full-stack app — React SPA, Express on Netlify Functions, Neon Postgres. Gamification on top.

20+API routes 8DB tables 100%serverless
Expand case study
Live · Full-stack · Serverless

TaskQuest Tracker

A gamified task tracker that turns daily routines into points, streaks, and rewards — shipped as a production full-stack app. React SPA on Netlify CDN, Express API bundled into a single serverless function, Neon Postgres for users, tasks, and sessions. Auth, guest mode, rewards shop, stats, and Google Calendar import — all solo.

The problem
To-do lists die without a feedback loop — and most side projects stop at the UI layer, never reaching production.
The approach
Gamification layer (points, streaks, rewards shop) on a real backend: Passport auth, persistent sessions, 20+ API routes across 8 DB tables.
The build
Monorepo — Vite + React frontend, Express API bundled to a 2.2 MB Netlify Function. Solved Neon serverless constraints, inline session table creation, and production build pipeline issues.
ReactExpressNeon PostgresNetlify FunctionsTypeScriptDrizzle ORM
Open the app ↗ Product case study ↗ Architecture one-pager ↗ GitHub ↗
+
more products in build — check back soon