AI Projects

Shipped AI products and tools — from personal experiments to autonomous systems for revenue and growth.

NextMove — AI-Powered Idea & Task Manager (iOS)

Personal iOS app that captures ideas, scores them using AI with RVS methodology, generates actionable tasks, and surfaces stale ideas for weekly review.

Business Problem

Ideas get lost without a structured system to capture, prioritize, and act on them. Most note apps dump everything into a flat list with no intelligence about what to do next.

Why AI vs Traditional

Gemini API scores each idea using RVS methodology and generates 3–6 concrete action steps — turning a raw thought into a prioritized, actionable plan without manual effort.

Architecture

Idea captured (collapsible card UI)
      ↓
Gemini API (RVS scoring + 3–6 action steps)
      ↓
Tasks linked to idea, recurring support
      ↓
Review queue surfaces stale/inactive ideas
      ↓
Capacitor (local storage, iOS native)

Implementation

Next.js 14 (App Router, static export) · Tailwind CSS · Zustand (localStorage persistence) · Capacitor 8 (iOS) · Google Gemini API

Business Value

A personal productivity system that turns captured ideas into scored, actionable tasks — built and shipped as a native iOS app. All data stays local; only idea metadata sent to Gemini.

ExpenseManager — AI-Powered iOS Expense Tracker

iOS app that parses natural language expense entries using Google Gemini.

Business Problem

Manual expense logging is tedious — users forget entries or input them inconsistently, making categories unreliable for reporting.

Why AI vs Traditional

Gemini extracts amount, category, date, and notes from a single sentence, handling varied phrasing without regex patterns or structured input forms.

Architecture

User types: "spent ₹450 on groceries"
      ↓
Gemini API (Flash / Flash Lite)
      ↓
{ amount: 450, category: "Groceries", date: today, note: "" }
      ↓
@capacitor/preferences (iOS UserDefaults)
      ↓
Recharts dashboard + searchable history

Implementation

React 18 + Vite · Capacitor 8 (iOS) · Google Gemini API (2.5 Flash / Flash Lite) · Recharts · @capacitor/preferences · Tailwind CSS + Framer Motion

Business Value

Shipped a production iOS app solo using AI-assisted development from idea to App Store.

Price Parity Monitor — Cross-Channel SKU Price Tracker

POC tool that crawls Amazon and Instamart product listings via Firecrawl and surfaces price parity violations against internal D2C prices — per SKU, across channels.

Business Problem

Price discrepancies across Amazon, Instamart, and D2C are common and go unnoticed without automated monitoring. Manual checks don't scale past a handful of SKUs.

Why AI vs Traditional

Firecrawl extracts price data from dynamically rendered ecommerce pages via a single API — no CSS selectors to maintain across Amazon, Instamart, and D2C layouts.

Architecture

User adds product URLs (Amazon, Instamart, internal D2C)
      ↓
Firecrawl API (AI-powered web extraction)
      ↓
{ price, title, availability } per channel
      ↓
Side-by-side price comparison dashboard
      ↓
Parity status: in-range / violation flagged

Implementation

Lovable (AI-assisted app builder) · Firecrawl API (AI web extraction) · React · Tailwind CSS

Business Value

Gives visibility into cross-channel price gaps per SKU without manual checks. Built and deployed as a working POC using Lovable + Firecrawl.

Internal Repo

Conversational Analytics — Natural Language Querying on Customer & Order Data

POC that takes plain English questions about customer behavior and order trends, translates them into SQL, executes the queries, and returns a structured summary answer.

Business Problem

Ad hoc questions about customer behavior and order trends require SQL knowledge or analyst time — creating a bottleneck for every stakeholder who wants to understand the data.

Why AI vs Traditional

An LLM translates a conversational question into one or more SQL queries, executes them against customer and order-level data, and synthesizes the results into a short, readable answer — no SQL required from the user.

Architecture

User asks: "Which cohort has the highest second-order rate this quarter?"
      ↓
LLM generates SQL query plan
      ↓
Executes queries against customer / order-level data
      ↓
LLM synthesizes query results
      ↓
Returns structured summary answer

Implementation

Python · LLM (in progress)

Business Value

Lets any stakeholder ask data questions in plain English and get a grounded, data-backed answer — without writing SQL or waiting for an analyst.

Internal Repo